Article Text

Download PDFPDF

Original article
The dialysis procedure as a trigger for atrial fibrillation: new insights in the development of atrial fibrillation in dialysis patients
  1. M S Buiten1,
  2. M K de Bie1,
  3. J I Rotmans2,
  4. B A Gabreëls3,
  5. W van Dorp4,
  6. R Wolterbeek5,
  7. S A Trines1,
  8. M J Schalij1,
  9. J W Jukema1,
  10. T J Rabelink2,
  11. L van Erven1
  1. 1Department of Cardiology, Leiden University Medical Center (LUMC), Leiden, The Netherlands
  2. 2Department of Nephrology, LUMC, Leiden, The Netherlands
  3. 3Department of Nephrology, Rijnland Hospital, Leiderdorp, The Netherlands
  4. 4Department of Nephrology, Kennemer Gasthuis, Haarlem, The Netherlands
  5. 5Department of Medical Statistics, LUMC, Leiden, The Netherlands
  1. Correspondence to Dr L van Erven, Department of Cardiology, Leiden University Medical Center, P O Box 9600, Leiden 2300 RC, The Netherlands; L.van_erven{at}


Aims Atrial fibrillation (AF) is common in dialysis patients and is associated with increased morbidity and mortality. The pathophysiology may be related to common risk factors for both AF and renal disease or to dialysis-specific factors. The purpose of this study was to determine whether and how AF onset relates to the dialysis procedure itself.

Methods All dialysis patients enrolled in the implantable cardioverter defibrillator-2 (ICD-2) trial until January 2012, who were implanted with an ICD, were included in this study. Using the ICD remote monitoring function, the exact time of onset of all AF episodes was registered. Subsequently, this was linked to the timing of dialysis procedures.

Results For the current study, a total of 40 patients were included, follow-up was 28±16 months, 80% male, 70±8 years old. A total of 428 episodes of AF were monitored in 14 patients. AF onset was more frequent on the days of haemodialysis (HD) (p<0.001) and specifically increased during the dialysis procedure itself (p=0.04). Patients with AF had a larger left atrium (p<0.001) and a higher systolic blood pressure before and after HD (p<0.001).

Conclusion This study provides insight in the exact timing of AF onset in relation to the dialysis procedure itself. In HD patients, AF occurred significantly more often on a dialysis day and especially during HD. These findings might help to elucidate some aspects of the pathophysiology of AF in dialysis patients and could facilitate early detection of AF in these high-risk patients.

  • Renal Disease
View Full Text

Statistics from


The incidence of cardiovascular disease in dialysis patients is high, with an annual cardiovascular mortality rate 5–30 times higher than the general population.13 Atrial fibrillation (AF), a major risk factor for ischaemic stroke and an independent predictor of death, is very common in dialysis patients.25 Furthermore, AF-related mortality in patients with end-stage renal disease (ESRD) is estimated to be twice as high as in patients without kidney disease.6 Despite our increasing knowledge about rhythm disorders, our knowledge on the relationship between AF and dialysis remains scarce.

In the general, non-dialysis population, risk factors for the development of AF include age, hypertension, obesity, diabetes, atrial enlargement and pre-existent cardiovascular disease.2 ,68 These risk factors are common in dialysis patients and could explain their high burden of AF. On the other hand, the dialysis procedure itself might provide unique risk factors for the onset of AF.8 ,9 Consequently, knowledge of the exact timing of AF onset in relation to the dialysis procedure itself could provide valuable information regarding the underlying pathophysiology.

The purpose of the present study was to determine the exact timing and characteristics of AF, in relation to the dialysis procedure in patients who received an implantable cardioverter defibrillator (ICD), as part of the ICD-2 study protocol.10 The remote monitoring function of the ICDs in these patients provided a unique possibility to thoroughly relate onset of AF to the dialysis procedure itself.


Study population

For this study, all patients currently enrolled in the ICD-2 trial (ISRCTN20479861), receiving a dual-chamber ICD (Biotronik, Berlin, Germany) between 2008 and 2012, were included. This ongoing randomised-controlled trial was designed to evaluate the effectiveness of ICDs in the prevention of sudden cardiac death in ESRD and the study protocol has been described previously.10 The study population consists of dialysis patients without a primary or secondary indication for an ICD. All patients provided informed consent, and the trial design was approved by the local ethics committee.

Implantable cardioverter defibrillator

All ICDs were implanted transvenously. The ICD was programmed to evaluate all p–p intervals. Registration of an atrial episode started if at least 36 out of 48 consecutive p–p intervals had a frequency higher than 180/min. An atrial episode was deemed terminated if 20 out of 24 consecutive p–p intervals had an AT/AF interval of <180/min. The atrial sensing threshold was programmed as 0.4 mV. Average p-wave sensing amplitude was calculated using the remote monitoring data. The implanted ICD contains an integrated antenna that transmits a heart rhythm registration and data on the timing of arrhythmias to an external remote monitoring device (Biotronik, Cardiomessenger 2-S), which subsequently sends the data to a central database every day.

Dialysis procedure

For this study, we included both patients on peritoneal dialysis (PD) and haemodialysis (HD). HD patients were dialysed 2–3 times a week with a dialysis time of 3–4 h. Ultrafiltration rate and dialysate potassium concentrations were based on target weight, actual body weight and serum potassium of the individual patients. PD patients received continuous dialysis (24 h a day, 7 days a week) in which the interval of intraperitoneal fluid replacement, which was performed at home, depended on the patient preferences and characteristics of the peritoneal membrane.

Atrial fibrillation

All AF episodes registered by the ICD were analysed using the intracardiac electrogram, stored in the central database. One of the authors (MSB) confirmed that the atrial episode registered by the ICD were actually AF, while blinded for the relationship of the atrial episode with the dialysis procedure. The date and time of AF onset were registered, as well as the episode duration. The days of the week were numbered such that the first day of the week, on which each individual patient underwent dialysis, was coded as day 1. Using this method, patients who performed dialysis on different days of the week were compared. For this particular analysis, we used patients performing HD three times a week, or performing PD.

Subsequently, AF onset was related to the start of the HD procedure. Onset of AF was categorised in the 7 h before dialysis, the 3–4 h during or the 7 h after HD. This window of 7 h was chosen because a larger time window would result in crossover into the next day.

Clinical variables

Baseline parameters were compared between patients with and without AF. As part of the ICD-2 protocol, all patients underwent two-dimensional transthoracic echocardiography (M3s probe, Vivid 7, GE Vingmed, Horton, Norway) at the time of inclusion. Left atrial (LA) diameter, left ventricular mass index (LVMI) and the presence of mitral valve regurgitation were assessed.11

If an HD patient developed AF on a dialysis day, the associated dialysis volume, potassium concentrations and blood pressures before and after the procedure were collected from an electronic patient database (Diamant, Diasoft, The Netherlands). Since serum potassium concentrations were only measured on a monthly basis, dialysate potassium concentration was used as an approximation of potassium shifts. For patients on HD, there is an inverse relation between the prescribed dialysate potassium concentration and the serum potassium concentration in order to remove sufficient amounts of potassium. Therefore, low potassium concentrations in the dialysate causes a high potassium shift from the serum during HD.

A single dialysis procedure from every single month of follow-up in patients without AF was randomly selected to compare dialysis procedures between patients with and without AF. In an effort to distinguish between patient-related and dialysis-related factors, additional data on HD procedures that were not associated with AF were collected from the patients who developed AF.


The primary outcome of this analysis was the onset of AF, and all episodes of AF were measured until the end of follow-up. If AF was registered by the ICD during follow-up, the treating nephrologist and cardiologist of that particular patient were informed and treatment was started at their discretion.


Continuous data were described by their mean (±SD) or as a median (25th–75th centile) if data were skewed and compared using an independent Student t test. Categorical data were presented as a percentage and compared using χ2 test or Fisher's exact probability test if appropriate.

All AF episodes were classified by the different days of the week (1–7). A repeated measures logistic regression (generalised estimating equations (GEE)) was used to compute the probability that AF developed on a dialysis day, thereby taking into account the within-patient correlation.

Subsequently, all AF episodes that occurred on a dialysis day were analysed. AF onset was categorised in the period before, during and after HD. Incidence rates between these periods were compared as rate-ratios using Poisson regression, again taking into account the within-patient repeated measures nature of the data. A GEE model with robust variance estimator was used.

The within-patient correlation was taken into account in both the analyses. However, since the number of AF episodes was not distributed evenly among patients, characteristics of patients who provided many AF episodes might still influence our outcomes. Therefore, both analyses were performed multiple times, each time excluding one single AF patient.

All statistical analyses were performed using SPSS (V.18.0, SPPS Inc. Chicago, Illinois, USA).



Forty patients were included in this study. Average follow-up was 28±16 months. Patients were predominantly male (80%), and the majority was treated with HD (65%). Average time on dialysis was 54±28 months, and average LVEF was 54%.

Patients on PD were younger (67±7 vs 72±7 years, p=0.045), had a smaller LA diameter (39±6 vs 44±6 mm, p=0.01) and a lower LVMI (105±28 vs 144±39, p=0.002) compared with HD patients.

Incidence of AF

In 14 patients, at least one episode of AF was recorded during follow-up (34%). Nine of these patients (64%) were not previously diagnosed with paroxysmal AF. Patients that developed AF had a significantly larger LA diameter (46±2 vs 40±6 mm, p<0.001). Further baseline characteristics did not differ between both groups (table 1).

Table 1

Baseline characteristics

In the 14 patients who developed AF, the ICD monitored a total of 10 196 days. In the HD group (n=11), a total of 8190 days were monitored, of which 1594 days (19%) were spent in AF (386 AF episodes). In the PD group (n=3), a total of 2006 days were monitored, including 64 days (3%) of AF (42 AF episodes). Median duration of the AF episodes was 212 min (42–441 min, table 2). Eighteen per cent of episodes lasted less than 6 min. The distribution of AF episodes ranged from 1 to 213 episodes per patient (table 2).

Table 2

Number of AF episodes per patient

AF in relation to the dialysis procedure

In the 11 HD patients, 10 received dialysis three times a week. One patient switched from HD twice a week to three times a week, and in this patient only the latter period was used. After AF onset was scored according to the days of the HD procedure, it was shown that the frequency of AF development on a dialysis day was approximately three times higher than on a non-dialysis day (282 vs 87episodes, p=0.001, figure 1A). No significant difference in AF onset between days of the week was observed in PD patients (figure 1B).

Figure 1

Onset of atrial fibrillation on different days of the week for haemodialysis(A) and peritoneal dialysis(B).

Out of the 296 AF episodes that occurred on a dialysis day, 236 episodes had both the start of the dialysis procedure and the exact timing of AF onset recorded. In the period between 7 h before and 7 h after HD, 205 AF episodes occurred. Eight per cent of these AF events commenced before the dialysis procedure, 48% began during HD and 43% of AF episode started afterwards (figure 2).

Figure 2

Onset of atrial fibrillation episodes in relation to the start of the dialysis procedure.

In the 7 h before the HD procedure, the rate of AF onset was approximately 0.003/person-hours compared with 0.04/person-hours during the HD procedure (rate ratio (RR) 0.075). Thus, the risk of developing AF during HD was approximately 13 times higher than before HD (p=0.03). The rate of AF onset in the 7 h after the HD procedure was 0.04/person-hours (RR 0.5), and the risk of developing AF during HD was approximately twice as high (p=0.001).

The distribution of AF onset around the HD procedure changed after excluding individual patients. After exclusion of patient #2 (table 2), the risk of developing AF during HD remained significantly higher than before HD (RR=0.5, p=0.001), but there was no statistical difference between the risk of AF onset during and after HD (RR=1, p=0.45). After exclusion of patient #7, the increased chance to develop AF during HD compared with before dialysis lost statistical significance (RR=0.06, p=0.07), whereas the chance to develop AF during HD remained two times higher than after HD (RR=0.5, p=0.001). Exclusion of the other patients had no significant effect. Furthermore, AF onset remained more frequent on a dialysis day compared with a non-dialysis day, after exclusion of individual patients.

Dialysis characteristics

Patients with AF developed AF episodes more frequently around HD procedures with a higher extracted volume (2120 vs 1611 mL, p<0.001) and a lower dialysate potassium concentration (1.6 vs 2.0 mEq/L, p<0.001). Furthermore, AF patients had a higher extracted volume during HD (2120 vs 1843, p≤0.01) and a lower dialysate potassium concentration (1.6 vs 1.9 mmol/L, p≤0.001) than patients who never developed AF (table 3).

Table 3

Dialysis characteristics


This study demonstrates a distinct relationship between AF and the HD procedure. AF onset was more frequent on the days of the HD procedure, and AF occurred mainly during the dialysis procedure itself. Furthermore, patients receiving PD showed significantly less episodes of AF. These findings might help to elucidate some aspects of the pathophysiology of AF in dialysis patients and might facilitate early diagnosis.

Although an association between HD and the onset of AF has previously been described, data on this relationship and its aetiology remain scarce.8 ,12 Ansari et al13 described that 9 out of a total of 10 monitored episodes of atrial arrhythmia started in the last hour of HD, based on patient complaints. A relationship between dialysis and arrhythmia was also described by Korzets et al,14 although no actual data supporting this claim were provided. Although the number of included patients was relatively small, the current analysis of 417 episodes of AF in 14 dialysis patients, based on continuous ICD telemonitoring, provides the largest analysis on the relationship between AF onset and the HD procedure currently available.

Dialysis-specific factors in the occurrence of AF

The current data show that the dialysis procedure itself is associated with the development of AF. Two pathways have been suggested as important in this relationship. The volume hypothesis postulates a major role for either volume overload or intravascular volume depletion in the development of AF, whereas the alternative pathway suggests an important role for electrolyte shifts in AF development.1517

The volume hypothesis

Volume overload causes atrial stretch, LV hypertrophy and neurohumeral alterations.17 ,18 This pathway would lead to a relatively high incidence of AF onset between dialysis days and just prior to the HD procedure, when fluid overload is at its largest. Since there was no increased incidence in the 7 h before the start of the HD procedure, volume overload seems to be an unlikely explanation for the increased incidence of AF in HD patients.

Previous studies have shown a profound drop in intracardiac diastolic and systolic pressure during HD.19 ,20 This suggests that the HD procedure might result in intravascular volume depletion, which inhibits the baroreflex, causing sympathetic activation and catecholamine release, which increases susceptibility to AF.21–,23 In the current study, AF onset occurred most frequently during the HD procedure and the incidence of AF onset increased continuously during HD. Furthermore, higher ultrafiltration volumes in dialysis sessions as well as lower diastolic pressure after dialysis were associated with the occurrence of AF. These factors suggest a role for intravascular volume depletion, in the development of AF in HD patients.

Electrolyte shifts

Trans-membranous fluxes of electrolytes, especially potassium, form another possible arrhythmogenic pathway.15 ,16 During HD, supraventricular ectopy and heart rate increase, whereas both intra-atrial conduction velocity and repolarisation duration decrease, all influenced by potassium concentration.15 ,24 ,25

We used dialysate potassium concentration as a surrogate for potassium flux. The development of AF was associated with a significantly lower potassium dialysate concentration, suggesting a possible role for potassium flux. However, if potassium flux would be the main culprit in the development of AF, one would expect a higher incidence of AF onset at the start of the HD procedure and a more even distribution of onset during the procedure than found in our data. In other words, the current data suggest an association with potassium concentration in dialysis fluid and AF development, but whether this is due to potassium flux is unclear.

An alternative explanation could be that low dialysate potassium concentrations lead to hypokalaemia at the end of HD, which subsequently causes arrhythmia. This would explain the gradual increase of AF onset during the HD procedure and the gradual decrease of AF onset in the hours after dialysis, when potassium redistributes and serum potassium concentration normalises. Unfortunately, we were not able to collect data on the actual serum potassium concentration before and after the HD session.

Peritoneal dialysis

Patients performing PD showed significantly less episodes of AF than HD patients, and the episodes did not appear to be related to the days of the week. This could be explained by the fact that PD is a dialysis modality in which solutes and water are being removed continuously. Therefore, electrolyte and volume shifts are more gradual compared with patients on HD. Furthermore, PD patients in the present population might represent a younger and healthier patient group. However, the number of PD patients in our analysis was too small to draw definite conclusions.

Clinical implications

Recent literature showed that atrial arrhythmias, as measured by an ICD, lasting for more than 6 min, were associated with an increased risk of ischaemic stroke and systemic embolism.26 In our data, AF duration was more than 6 min in 82% of the cases, thus implicating an increased thromboembolic risk. Stroke prevention is difficult in HD patients since ESRD is associated with an elevated bleeding risk as well.27 Therefore, the current opinion is to refrain from routinely using oral anticoagulants as primary prevention of stroke in HD patients and to evaluate on a single-patient basis.28 This evaluation might be facilitated by a 12-lead ECG at the end of dialysis sessions at a regular basis, especially since 64% of the patients with AF found in our analysis were newly diagnosed.

The possible role of intravascular volume depletion as an instigator for AF might suggest a benefit for more frequent dialysis procedures or the recommendation of lower dietary fluid intake. Frequent HD procedures (5–6 times a week) have already been shown to decrease mortality and LV mass with a positive effect on electrolyte metabolism.29

Further studies on the association between AF and the dialysis procedure are needed. More insights in the association between serum potassium concentration around the time of dialysis and AF development might benefit future AF prevention. Furthermore, novel diagnostic tools such as the body composition monitor might be useful to thoroughly relate fluid state and AF development in ESRD.30


This study is limited by the relatively small patient population, which is partly compensated by the large number of AF episodes monitored. Despite remote monitoring, AF episodes may be missed due to undersensing or overwriting of earlier episodes, by the remote monitoring system. The aim of this study was to analyse AF onset in relation to the dialysis procedure, we did not have a sufficient number of patients to be able to evaluate the clinical consequences of the presence of AF, in our patient group. Furthermore, data on potassium levels before and after the dialysis procedure were not available.


This study provides insight into the exact timing of AF onset in relation to the dialysis procedure itself. In HD patients, AF occurred significantly more often on a dialysis day and especially during HD. These findings might help to elucidate some aspects of the pathophysiology of AF in dialysis patients and could facilitate early detection of AF in these high-risk patients.

Key messages

What is already known on this subject?

  • Atrial fibrillation (AF) is common in dialysis patients, and AF-related mortality in these patients is much higher than in patients without kidney disease. An association between haemodialysis (HD) and the onset of AF has previously been described in a very small number of AF episodes, but an accurate analysis of the association between dialysis and AF is not available.

What this study adds?

  • Although the number of investigated patients remains relatively small, the present study, based on continuous ICD telemonitoring, provides the largest and most accurate analysis on the relationship between AF and the dialysis procedure currently available. In HD patients, AF occurred significantly more often on a dialysis day and especially during the HD procedure.

How might this impact on clinical practice?

  • Information on the development of arrhythmias in dialysis patients is much needed given the disproportionally high cardiovascular mortality in these patients. These findings might help to elucidate some aspects of the pathophysiology of AF in dialysis patients. Furthermore, stroke prevention in dialysis patients with AF is currently a topic of much discussion and information on the onset of AF might facilitate early diagnosis and treatment decisions.


View Abstract


  • Contributors All authors contributed equally to the writing of this manuscript.

  • Funding The Department of Cardiology receives unrestricted grants from Biotronik (Berlin, Germany), Boston Scientific (Natick, Massachusetts), Medtronic (Minneapolis, Minnesota). The ICD2 study is supported by an unrestricted educational research grant from Biotronik (Berlin Germany). There is no further involvement of any company with this article in any other way.

  • Competing interests None.

  • Patient consent Obtained.

  • Ethics approval Commissie Medischie Ethiek Leiden, the Netherlands.

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Linked Articles