International Journal of Cardiology 186 (2015) 170–177

Contents lists available at ScienceDirect

International Journal of Cardiology

j ourna l homepage: www.e lsev ie r .com/ locate / i j ca rd

Mortality, sudden death and indication for cardioverter defibrillatorimplantation in a dialysis population

Simonetta Genovesi a,c,⁎, Luca Porcu b, Maria Carmen Luise c, Hilary Riva c, Elisa Nava c, Andrea Stella a,c,Claudio Pozzi d, Patrizia Ondei e, Claudio Minoretti f, Maurizio Gallieni g, Giuseppe Pontoriero h,Ferruccio Conte i, Valter Torri b, Antonio Vincenti j

a Nephrology Unit, San Gerardo Hospital, Monza, Italyb Laboratory of Methodology for Biomedical Research, Oncology Department, IRCCS — Istituto di Ricerche Farmacologiche Mario Negri, Milan, Italyc Department of Health Sciences, University of Milan-Bicocca, Italyd Nephrology Unit, Bassini Hospital, Cinisello Balsamo, Italye Nephrology Unit, Ospedali Riuniti Hospital, Bergamo, Italyf Nephrology Unit, Sant'Anna Hospital, Como, Italyg Nephrology Unit, San Carlo Borromeo Hospital, Milan, Italyh Nephrology Unit, Alessandro Manzoni Hospital, Lecco, Italyi Nephrology Unit, Uboldo Hospital, Cernusco sul Naviglio, Italyj Electrophysiology and Cardiac Pacing Unit, San Gerardo Hospital, Monza, Italy

⁎ Corresponding author at: Department of Health SBicocca, Via Cadore 48, 20900 Monza, Italy.

E-mail address: simonetta.genovesi@unimib.it (S. Gen

http://dx.doi.org/10.1016/j.ijcard.2015.03.1780167-5273/© 2015 Elsevier Ireland Ltd. All rights reserved

a b s t r a c t

a r t i c l e i n f o

Article history:

Received 24 October 2014Received in revised form 9 February 2015Accepted 16 March 2015Available online 17 March 2015

Keywords:Implantable cardioverter defibrillatorEnd stage renal diseaseMortalitySudden deathHemodialysis

Background: The incidence of sudden death among dialysis patients is high, but end stage renal disease was anexclusion criterion in the trials that demonstrated the benefit of implantable cardioverter defibrillator (ICD)for sudden death prevention.Methods:Dialysis patients alive on January 2010or starting dialysis between January 2010 and January 2013wereenrolled and retrospectively evaluated. Patientswere divided into three groups: No-Indication, Indication–With ICDand Indication–Without ICD. Cox and Fine and Gray regression models were used to estimate the total and cause-specific (sudden or non-sudden) mortality hazard ratio (HR, HRcpRisk), respectively. Survival was defined as thetime from start of dialysis to the time of death.Results: 154/2072 patients (7.4%) had indication for ICD implantation and 52 (33.8%) of them received the device;688 (33.2%) deaths were recorded. Mortality was different among groups [Indication–With ICD vs No-Indication:HR 1.59 (95% CI 1.06–2.38) and Indication–Without ICD vs No-Indication: HR 2.67 (95% CI 2.09–3.39, p b 0.001)].

84/688 (12.2%) were sudden deaths. The cumulative incidence of sudden death was higher in patients with ICD in-dication [Indication–With ICD vs No-Indication HRcpRisk 3.21 (95% CI 1.38–7.40) and Indication–Without ICD vs No-Indication: HRcpRisk 4.19 (95% CI 2.38–7.39), p b 0.001], but alsoNo-Indication patients showed a high rate of suddendeath [8.5% (95% CI.6.5–10.9) at 8 years of follow-up].Conclusions:Dialysis patientswith ICD indicationhad aworse survival thanNo-Indication subjects and theprognosiswas particularly poor for the Indication–Without ICD group. Sudden death incidence was much higher than in thegeneral population, even among No-Indication subjects.

© 2015 Elsevier Ireland Ltd. All rights reserved.

1. Introduction

The incidence of sudden death among patients with end stage renaldisease (ESRD) undergoing hemodialysis (HD) or peritoneal dialysis ishigh. In a European HD population about 19% of deaths were due tosudden death [1], while the 2013 US register reported that 27% ofESRD patients died suddenly [2]. In the USA 49/1000 HD patients/yearand 36/1000 peritoneal dialysis patients/year die from sudden death,

ciences, University of Milano

ovesi).

.

but it is not known how many dialysis patients have an indication forimplantable cardioverter defibrillator (ICD) according to the cardiologicguidelines. Only 5/1000 patients/year receive an ICD [2]. There are sev-eral reasons for the underuse of the device in the dialysis population.First, the presence of ESRD was an exclusion criterion in the major car-diologic trials that demonstrated that the ICD has a survival benefit inhigh risk populations [3–5] and for this reason the usefulness of ICD isnot yet established in dialysis patients. However, the most recent ICDimplantation guidelines do not exclude the possibility to receive the de-vice for a dialysis patient presenting the indication, suggesting howeverto pay particular attention while making the decision [6]. Second, in thepresence of chronic kidney disease or ESRD, the overall mortality of ICD

Table 1Demographic and clinical patient characteristics.

Patient group

No-Indication(N = 1918)

Indication–Without ICD(N = 119)

Indication–With ICD(N = 52)

Patient characteristicsAge (years) at the start of dialysis N 1918 119 52

Median 68.2 71.8 69.4Range 12.9–94.4 18.5–89.8 46.2–85.2

GenderFemale N (%) 739 (38.5) 30 (25.2) 11 (21.2)Male 1179 (61.5) 89 (74.8) 41 (78.8)

NYHA class2 187 (9.8) 37 (31.1) 12 (23.1)3 60 (3.2) 27 (22.7) 16 (30.8)4 0 (0.0) 8 (6.7) 3 (5.8)

Left ventricular ejection fractionb35% N (%) 0 (0.0) 84 (70.6) 37 (71.2)≥35% 1918 (100.0) 35 (29.4) 15 (28.8)

Type of dialysisHemodialysis N (%) 1693 (88.3) 104 (87.4) 42 (80.8)Peritoneal dialysis 225 (11.7) 15 (12.6) 10 (19.2)

ComorbiditiesIschemic heart disease

Yes N (%) 610 (31.8) 87 (73.1) 37 (71.2)Diabetes mellitus

Yes N (%) 499 (26.0) 47 (39.5) 19 (36.5)Atrial fibrillation

Yes N (%) 473 (24.7) 56 (47.1) 26 (50.0)

Fig. 1. Survival curves of the three study groups. Scenario A: Patients with implanted ICDafter the start of dialysis are classified as ‘Indication–Without ICD’ before ICDimplantation. Scenario B: Patients with implanted ICD after the start of dialysis areclassified as ‘No-Indication’ before ICD implantation.

171S. Genovesi et al. / International Journal of Cardiology 186 (2015) 170–177

patients is significantly higher than that reported in patients with ICDand normal kidney function and the risk of death after ICD implantationfor primary prevention is proportional to chronic kidney disease sever-ity [7–14]. Third, the infective and implant-procedure-related complica-tions in chronic kidney disease patients are higher than in the non-chronic kidney disease population [15]. There are very few mortalitystudies in dialysis patients with ICD indication comparing ICD patientswith no-ICD patients and the results are controversial [16]. Moreoverthere are no studies comparing total mortality and sudden death inci-dence in dialysis patients with and without ICD indication.

The aim of this study was to estimate, in a dialysis population, theprevalence of patients with ICD indication and to compare total andsudden mortality occurring in this subgroup with those of subjectswithout ICD indication. In addition the prognosis of patients with ICDindication, according to ICD presence, has been investigated.

2. Methods

This is an Italian multicenter retrospective study.All dialysis patients (undergoing HD or peritoneal dialysis), alive on

the 1st of January 2010 or starting dialysis between the 1st January 2010and the 31st of January 2013 (recruitment time), were enrolled andtheir clinical charts revised. For all patients the ICD indication was eval-uated; on the 31st of January 2013 for alive subjects and at the time ofdeath if deceased.

Primary prevention indication was defined by MADIT I and II and/orSCDHeFT criteria [3–5], while secondary prevention was considered inpatients who had a history of cardiac-arrest.

Patientswere considered eligible for the study only if an echocardio-gram with a measured value of left ventricular ejection fraction, madewithin 6 months before recruitment if alive, or 6 months before deathif deceased, was available. Patients were divided into three groups:without ICD indication (No-Indication), with ICD indication and withimplanted ICD (Indication–With ICD), andwith ICD indication butwith-out implanted ICD (Indication–Without ICD). Because the exact time ofICD indication starting was not available, patients having ICD implanta-tion after the start of dialysis were analyzed in a doublemanner: classify-ing them as Indication–Without ICD (scenario A) or as No-Indication

172 S. Genovesi et al. / International Journal of Cardiology 186 (2015) 170–177

group (scenario B) from the start of dialysis to ICD implantation. At thedate of ICD implantation, patients shifted from Indication–Without ICDto Indication–With ICD in case of scenarioA, and fromNo-Indication to In-dication–With ICD group in case of scenario B.

In all groups of patients, the presence of the following comorbiditieswas collected: ischemic cardiac disease, diabetes mellitus and atrialfibrillation. Death causes were derived from medical records. Suddendeath was defined as spontaneous death preceded by a sudden loss ofconsciousness within 1 h after onset of acute symptoms, even in thepresence of pre-existing heart disease, but with unexpected timingand mode. Nephrologists or relatives were interviewed to confirm allcases of sudden death. Procedures were performed according to theHelsinki declaration for ethics treatment of human subjects and ap-proved by the local ethical committee.

2.1. Statistical analysis

Study endpoints were:

a. prevalence of ICD indication defined as the proportion of dialysis pa-tients having ICD indication at the recruitment time

b. prevalence of ICD treated patients defined as the proportion of dial-ysis patients having ICD at the recruitment time

c. overall survival defined as the time from the start of dialysis to thetime of death from any cause

d. cause of death (sudden or no-sudden).

In survival analysis dialysis patients alive on the 1st of January 2010contributed to the population at risk of death only after the 1st ofJanuary 2010; in order to account for the fact that patients became atrisk after the interval time between the beginning of dialysis and 1stof January 2010 (left-truncation or delayed entry) [17]. Because ofleft-truncation and patients shifting from Indication–Without ICDgroup population to the Indication–With ICD group population in sce-nario B, survival functions were non-parametrically estimated usingthe Simon–Makuch estimator [18] and statistically compared usingthe Mantel–Byar test [19]; Cox regression model was used to estimatethe hazard ratio (HR).

In order to estimate the statistical association between the threepopulations at risk of death and the specific cause of death and to

Table 2Univariate and multivariate Cox analysis results on death risk for any cause in the total popula

Analysis Variable Category

Univariate Patient characteristicsAge at start of dialysis (years) –

Gender FemaleMale

Left ventricular ejection fraction (%) b35≥35

Ischemic heart disease NoYes

Atrial fibrillation NoYes

Diabetes mellitus NoYes

TreatmentType of dialysis Peritoneal dialysi

HemodialysisMultivariate Age at start of dialysis (years) –

Left ventricular ejection fraction (%) b35≥35

Ischemic heart disease NoYes

Atrial fibrillation NoYes

Diabetes mellitus NoYes

identify patient characteristics and treatment statistically associated tothe specific cause of death, a survival analysis in thepresence of compet-ing risks was performed. The cumulative incidence function for failureof sudden death and other causes of death were computed using Stata'sstcompet command. The Fine andGray regressionmodelwas used to es-timate the cause-specific hazard ratio (HRcpRisk); HRcpRiskwas computedusing Stata's stcrreg command.

Survival status was updated on the 31st of January 2013 for the No-Indication group and on the 31st of January 2014 for the Indication–With ICD and Indication–Without ICD groups; median follow-up wasestimated using the inverse Simon–Makuch method, and in order toquantify the completeness of follow-up at the update of survival statusa completeness index was used.

Baseline covariate distributions were summarized using descriptivestatistics (median and range for continuous variables, and absolute andpercentage frequencies for categorical variables).

Because of the descriptive nature of this study, hypothesis testingwas applied qualitatively and not formally (e.g. no threshold for statis-tical significance level was defined).

Statistical analysis was performed using Stata software, version 12.1(StataCorp. 2011. Stata Statistical Software: Release 12. College Station,TX: StataCorp LP).

3. Results

Two thousand and seventy two patientswere enrolled from7 north-ern Italian centers. Demographic and clinical patient characteristics areshown in Table 1.

The prevalence of patients with ICD indication was 7.4% (154/2072,95% CI: 6.3–8.7%) and only 33.8% (52/154, 95% CI: 26.3–41.8%) of themhad received the device by the 31st of January 2013. Seventeen patientsshifted from Indication–Without ICD to Indication–With ICD group dur-ing follow-up. The ICD indicationwas establishedwithMADIT criteria in76 (49.3%) patients, with SCDHeFT criteria in 53 (34.4%), while 25(16.2%) had a secondary prevention indication.

3.1. Survival analysis

The median follow-up was 1.7 years (range 0.01–40.9), for No-Indication, 2.8 years (range 1.0–20.0) for Indication–With ICD and3.6 years (range 0.2–41.1) for Indication–Without ICD: 586/1918

tion.

Hazard ratio 95% CI Z p-Value

1.05 1.05–1.06 13.29 b0.0011 – −0.07 0.9450.99 0.85–1.161 – −6.04 b0.0010.47 0.37–0.601 – 4.44 b0.0011.41 1.21–1.641 – 5.74 b0.0011.59 1.36–1.861 1.28–1.89 4.11 b0.0011.41 1.20–1.66

s 1 – 3.12 0.0021.52 1.17–1.991.05 1.05–1.06 12.56 b0.0011 – −4.38 b0.0010.57 0.44–0.731 – 1.40 0.1631.12 0.95–1.321 – 1.49 0.1371.13 0.96–1.331 – 3.33 0.0011.33 1.12–1.57

173S. Genovesi et al. / International Journal of Cardiology 186 (2015) 170–177

(30.5%) patients died in No-Indication, 25/52 (48.1%) in Indication–With ICD and 77/119 (64.7%) in the Indication–Without ICD group.Completeness of follow-up was 99.1% for No-Indication, 97.6% for Indi-cation–With ICD and 98.9% for the Indication–Without ICD.

Survival curves of the three groups of patients are shown in Fig. 1: themedian survival was 3.7 years (95% CI 3.4–4.3) in No-Indication, 2.0 years(95% CI 0.9–3.5) in Indication–With ICD, and 1.2 years (95% CI 0.7–1.7) inIndication–Without ICD (scenario A) and 3.7 years (95% CI: 3.4–4.3) inNo-Indication, 2.0 years (95% CI: 0.9–3.5) in Indication–With ICD,0.8 years (95% CI: 0.4–1.4) in Indication–Without ICD (scenario B).

In the three groups the mortality was significantly different(p b 0.001) in both scenario A [Indication–With ICD vs No-Indication:HR 1.58 (95% CI 1.05–2.36) and Indication–Without ICD vs No-Indication: HR 2.31 (95% CI 1.82–2.95)], and scenario B [Indica-tion–With ICD vs No-Indication: HR 1.59 (95% CI 1.06–2.38) andIndication–Without ICD vs No-Indication: HR 2.67 (95% CI 2.09–3.39)].

The mortality was significantly lower in Indication–WithICD compared to the Indication–Without ICD group in scenario B(χ21df,Mantel–Byar = 4.30; p = 0.038), while a not remarkable differencewas detected in scenario A (χ21df,Mantel–Byar = 2.15; p = 0.143).

The survivalmultivariate Cox analysis of the three groups of patientsshowed that factors associated with increased mortality were older ageat start of dialysis therapy (HR 1.05, 95% CI 1.05–1.06, p b 0.001), leftventricular ejection fraction lower than 35% (HR 0.57, 95% CI 0.44–0.73 p b 0.001) and the presence of diabetes mellitus (HR 1.33, 95% CI1.12–1.57, p = 0.001) (Table 2). When dialysis modality was added tothe model, results did not change and HD patients showed a higherrisk of mortality than peritoneal dialysis patients (HR 1.41, 95% CI1.08–1.85, p = 0.011, data not shown).

The survival analysis performed only in patients with ICD indication,are shown in Table 3: adjusted HRs of 1.34 (95% CI: 0.84–2.13, p =0.220) in scenario A and 1.51 (95% CI: 0.95–2.40, p= 0.083) in scenarioB were estimated for the Indication–Without ICD compared to the

Table 3Univariate and multivariate Cox analysis results on death risk for any cause in Indication–With

Scenario Analysis Variable Category

A Univariate ICD IndicatioIndicatio

Multivariate ICD IndicatioIndicatio

Patient characteristicsAge (years)Left ventricular ejection fraction (%)Ischemic heart disease No

YesAtrial fibrillation No

YesDiabetes mellitus No

YesTreatmentType of dialysis Peritone

HemodiaB Univariate ICD Indicatio

IndicatioMultivariate ICD Indicatio

IndicatioPatient characteristicsAge (years)Left ventricular ejection fraction (%)Ischemic heart disease No

YesAtrial fibrillation No

YesDiabetes mellitus No

YesTreatmentType of dialysis Peritone

Hemodia

Indication–With ICD group. A higher mortality was associated withage in scenario A (HR 1.04, 95%CI 101–1.07, p = 0.013), and the resultwas substantially confirmed in scenario B (HR: 1.03, 95%CI: 1.00–1.06,p = 0.091).

3.2. Sudden death

During the follow-up 84/688 (12.2%) sudden deaths were recorded:61/586 (10.4%) in No-Indication, 6/25 (24.0%) in Indication–With ICDand 17/77 (22.1%) in the Indication–Without ICD group.

In Fig. 2 the cumulative incidence of sudden death in the follow-upperiod is shown. The cumulative incidence of sudden death at 8 yearswas 8.5% (95% CI 6.5–10.9) among No-Indication patients and 23.3%(95% CI 9.4–40.9) and 21.1% (95% CI 12.3–31.5) in Indication–WithICD and the Indication–Without ICD patients, respectively. Suddendeath incidence was significantly higher in patients with ICD indication(p b 0.001) in both scenario A [Indication–With ICD vs No-Indication:HRcpRisk 3.19 (95% CI 1.37–7.45) and Indication–Without ICD vs No-Indication: HRcpRisk 3.85 (95% CI 2.20–6.76)] and scenario B [Indication–With ICD vs No-Indication: HRcpRisk 3.21 (95% CI 1.38–7.40) andIndication–Without ICD vs No-Indication: HRcpRisk 4.19 (95% CI2.38–7.39)].

Independent predictors of sudden death in the total populationwereolder age at the start of dialysis therapy (HRcpRisk 1.02, 95% CI 1.00–1.04,p = 0.028), left ventricular ejection fraction lower than 35% (HRcpRisk

0.45, 95% CI 0.24–0.81, p= 0.009) and the presence of ischemic cardiacdisease (HRcpRisk 1.83, 95% CI 1.15–2.93, p = 0.012) and, lightly, diabe-tesmellitus (HRcpRisk 1.56, 95%CI 0.99–2.46, p=0.056) (Table 4).Whenthe analysis was done only in patients from the No-Indication group, di-abetes mellitus remained the only variable strongly associated withsudden death (HRcpRisk 2.08, 95% CI 0.1.23–3.52, p = 0.006) and ische-mic cardiac disease had a borderline significance (HRcpRisk 1.65, 95% CI0.98–2.77, p = 0.060) (Table 5).

ICD and Indication–Without ICD groups.

Hazard ratio 95% CI Z p-Value

n–With-ICD 1 – 1.46 0.144n–Without-ICD 1.41 0.89–2.23n–With-ICD 1 – 1.23 0.220n–Without-ICD 1.34 0.84–2.13

1.04 1.01–1.07 2.47 0.0131.00 0.97–1.03 −0.02 0.9841 – 0.79 0.4301.22 0.74–2.011 – −0.75 0.4520.85 0.56–1.291 – 1.00 0.3171.24 0.81–1.90

al dialysis 1 – 0.52 0.606lysis 1.17 0.65–2.09n–With-ICD 1 – 2.06 0.040n–Without-ICD 1.62 1.02–2.56n–With-ICD 1 – 1.73 0.083n–Without-ICD 1.51 0.95–2.40

1.03 1.00–1.06 1.69 0.0911.00 0.98–1.03 0.12 0.9021 – 0.54 0.5901.15 0.70–1.891 – 0.03 0.9731.01 0.66–1.551 – 0.70 0.4831.17 0.76–1.80

al dialysis 1 – 0.71 0.477lysis 1.24 0.69–2.22

Fig. 2. Cumulative incidence of sudden death and of death due to other causes. Scenario A: Patients with implanted ICD after the start of dialysis are classified as ‘Indication–Without ICD’before ICD implantation. Scenario B: Patients with implanted ICD after the start of dialysis are classified as ‘No-Indication’ before ICD implantation.

174 S. Genovesi et al. / International Journal of Cardiology 186 (2015) 170–177

Sudden death incidence was not remarkably different in the twogroups of patients with ICD indication [Indication–Without ICD vs Indi-cation–With ICD: HRcpRisk 1.20 (95%CI 0.46–3.13), p = 0.7 in scenario Aand Indication–Without ICD vs Indication–With ICD: HRcpRisk 1.29(95%CI 0.49–3.39), p= 0.6 in scenario B]. At the beginning of the obser-vation period there was a substantial difference in the annual rate ofsudden death between Indication–With ICD and Indication–WithoutICD patients [4.0% (95% CI 0.3–16.9) vs 9.1% (95% CI 3.3–18.4), 7.3%(95% CI 01.3–20.7) vs 13.9% (95% CI 6.5–24.1) and 10.6% (95% CI 2.7–24.8) vs 16.3% (95% CI 8.3–26.7) at 1, 2 and 3 years of follow-up, respec-tively], which was gradually reduced until it disappeared during thefollow-up (Fig. 3).

Twelve/254 (4.7%) peritoneal dialysis patients and 72/1872 (3.8%)HD patients suffered sudden death. In 48 HD patients the timing ofdeath with respect to the HD session was established: 22/48 (45.8%)sudden deaths happened during the first inter-dialytic interval and18/48 (37.5%) during the last long inter-dialytic interval of the week.Three/48 patients who died suddenly were ICD patients.

4. Discussion

In our population of dialysis patients, the prevalence of subjectswithan indication for ICD implantation was about 7%, but only a minority ofthem actually received the device. From our data the criterion bywhicha patient is considered eligible to receive the ICD it is not clear: in fact,there are no important differences between patients with or withoutthe device. Mortality rate is high in the study population, particularlyin the first year after the start of dialysis, but at least half of the patientsat the beginning of renal replacement therapy have a life expectancythat is greater than one year andwould therefore be suitable for ICD im-plantation, if an indication was present [6]. As expected, older age anddiabetes mellitus are independent predictors of death. As in the generalpopulation, low ventricular ejection fraction is associated with in-creased total mortality.

The comparison among the three groups of study patients (No-Indication, Indication–With ICD and Indication–Without ICD), clear-ly shows that patients with an indication for ICD implantation have

Table 4Univariate and multivariate Fine and Gray analysis results on death risk for sudden and not sudden deaths in the total population.

Analysis Variable Category Sudden death Other causes

Hazard ratio 95% CI Z p-Value Hazard ratio 95% CI Z p-Value

Univariate Patient characteristicsAge at start of dialysis (years) – 1.02 1.01–1.04 3.05 0.002 1.05 1.04–1.06 3.05 b0.001Gender Female 1 – 0.78 0.433 1 – −0.51 0.608

Male 1.20 0.76–1.89 0.96 0.81–1.13Left ventricular ejection fraction (%) b35 1 – −4.05 b0.001 1 – −3.47 0.001

≥35 0.31 0.18–0.55 0.60 0.45–0.80Ischemic heart disease No 1 – 3.99 b0.001 1 – 2.48 0.013

Yes 2.41 1.57–3.73 1.23 1.04–1.45Atrial fibrillation No 1 – 1.19 0.235 1 – 5.16 b0.001

Yes 1.32 0.84–2.07 1.56 1.32–1.85Diabetes mellitus No 1 2.93 0.003 1 – 2.42 0.016

Yes 1.92 1.24–2.97 1.25 1.04–1.49TreatmentType of dialysis Peritoneal dialysis 1 – −0.69 0.489 1 – 3.63 b0.001

Hemodialysis 0.81 0.45–1.47 1.74 1.29–2.34Multivariate Age at start of dialysis (years) – 1.02 1.00–1.04 2.19 0.028 1.05 1.04–1.06 11.29 b0.001

Left ventricular ejection fraction (%) b35 1 – −2.63 0.009 1 – −2.14 0.033≥35 0.45 0.24–0.81 0.72 0.53–0.97

Ischemic heart disease No 1 – 2.53 0.012 1 – 0.10 0.919Yes 1.83 1.15–2.93 1.01 0.85–1.20

Atrial fibrillation No 1 – −0.002 0.998 1 – 1.70 0.088Yes 1.00 0.96–1.33 1.17 0.98–1.39

Diabetes mellitus No 1 – 1.91 0.056 1 – 2.01 0.045Yes 1.56 0.99–2.46 1.21 1.00–1.45

175S. Genovesi et al. / International Journal of Cardiology 186 (2015) 170–177

a worse prognosis than those without. The survival curves also showa particularly poor trend for the Indication–Without ICD group andin these patients the risk of mortality is greater than in implantedpatients. Evidence from literature reports that ICD patients withchronic kidney disease or ESRD have a higher mortality comparedto those with the device and preserved renal function [7–14]. Theonly randomized trial focused on the usefulness of ICD implantationfor primary prevention in dialysis patients has not yet provided re-sults [20] and only four observational and retrospective studiesevaluated the benefit of ICD within populations of ESRD patients.One of these, based on a large database, demonstrates a clear benefitof the ICD in terms of total mortality, but it includes only secondaryprevention patients [21]. Two studies, involving primary preventionpatients, arrive at two opposite conclusions: in one of the studies, animproved survival in the group of ESRD patients receiving ICD is ob-served [22], while in the other the mortality is similar in HD patients

Table 5Univariate and multivariate Fine and Gray analysis results on death risk for sudden and not su

Analysis Variable Category Sudden death

Hazard ratio

Univariate Patient characteristicsAge at start of dialysis (years) – 1.02Gender Female 1

Male 1.08Ischemic heart disease No 1

Yes 1.97Atrial fibrillation No 1

Yes 1.33Diabetes mellitus No 1

Yes 2.35TreatmentType of dialysis Peritoneal dialysis 1

Hemodialysis 1.11Multivariate Age at start of dialysis (years) – 1.02

Ischemic heart disease No 1Yes 1.65

Atrial fibrillation No 1Yes 1.15

Diabetes mellitus No 1Yes 2.08

with and without the device [23]. Moreover, a recent matched co-hort study, performed by Pun et al., failed to demonstrate a signifi-cant association between primary prevention ICDs and reducedmortality among ESRD patients receiving dialysis [24]. In all thesestudies, however, there is a methodological problem: control pa-tients (i.e. without ICD) are recruited only on the basis of havingan ejection fraction b 35% and their follow-up begins in a momentof their clinical history chosen in an arbitrary manner. In ourstudy, not being able to accurately identify the time of occurrenceof the indication for ICD implantation, we began our observationfrom the time of the start of dialysis therapy. This was done becausebeing on dialysis was the unifying factor for the three groups ofpatients and the start of renal replacement therapy representedthe beginning of exposure to an additional risk factor of total andsudden mortality, beyond those traditionally known. Our resultssuggest that ICD implantation may have some benefit in terms of

dden deaths in the No-Indication group.

Other causes

95% CI Z p-Value Hazard ratio 95% CI Z p-Value

1.00–1.04 2.46 0.014 1.06 1.05–1.06 11.85 b0.001– 0.29 0.775 1 – −0.79 0.4280.64–1.81 0.93 0.78–1.11– 2.66 0.008 1 – 1.44 0.1501.20–3.26 1.14 0.95–1.37– 1.04 0.297 1 – 4.77 b0.0010.78–2.27 1.57 1.30–1.89

3.28 0.001 1 – 1.84 0.0661.41–3.92 1.20 0.99–1.46

– 0.27 0.787 1 – 3.52 b0.0010.51–2.45 1.81 1.30–2.520.99–1.03 1.64 0.101 1.05 1.04–1.06 11.34 b0.001– 1.88 0.060 1 – −0.02 0.9810.98–2.77 1.00 0.83–1.20– 0.49 0.621 1 – 1.83 0.0680.66–2.03 1.19 0.99–1.44– 2.73 0.006 1 – 1.50 0.1331.23–3.52 1.16 0.95–1.42

Fig. 3. Annual rate of sudden death in the three study groups.

176 S. Genovesi et al. / International Journal of Cardiology 186 (2015) 170–177

survival in ESRD patients, but open a scenario much wider asregards the sudden mortality in this population. In the previousstudies performed in HD-ICD patients, only total, but no suddenmortality was considered. In our population the cumulative inci-dence of sudden death is high and represents 12% of all causes ofdeath, but with a completely different ratio from that observed inthe general population [25]. Subjects with indication for ICD im-plantation more frequently die of sudden death compared withthose without indication, but what clearly emerges is that in ourpopulation the incidence of sudden death is extremely high evenin patients with preserved ejection fraction (i.e. No-Indicationgroup). Patients with left ventricular ejection fraction lower than35% and diabetic patients die more frequently of sudden death.Moreover, the presence of coronary disease is associated with an in-creased risk of sudden mortality, independently from left ventricu-lar ejection fraction value. Our data do not show a reduction in theincidence of sudden death in patients with ICD after 8 years offollow-up, while there is a more favorable trend in implanted pa-tients in the early dialysis years (Fig. 3). These observations suggestthat the ESRD “per se” is a risk factor of sudden death. It has beenshown that, in dialysis patients, diabetes mellitus and atrial fibrilla-tion, are predictors of sudden death [1]. Our data confirm that thesudden death risk is higher in HD diabetic patients and this is partic-ularly true for subjects with preserved left ventricular ejection frac-tion. It has been suggested that in the ESRD population a highcardiac mass constitutes a risk factor for sudden death that is impor-tant beyond a reduced systolic function [26]. Moreover, even non-traditional risk factors such as high degree of inflammation andlow values of vitamin D have been associated with an increased in-cidence of sudden death in ESRD patients [27,28]. Many of the fac-tors mentioned above are frequently present in patients on renalreplacement therapy, beyond treatment modality (HD or peritonealdialysis), however some additional remarks should be made for HDpatients. Several authors described a relationship between thetiming of sudden death and HD session [29,1] and we confirm thisobservation: the first short interdialytic interval and the end of thelong interval are the moments of the week when most frequentlyour HD patients run the risk of experiencing sudden death. In the firstcase, the patient suffers a sudden and significant decrease in plasmapotassium concentration, in the second he presents hyperkalemia andacidosis. Both conditions can lead to electrical instability of theheart cells, which could lead to different types of life-threatening ar-rhythmias (tachy or brady-arrhythmias). Low potassium and calciumconcentrations in the dialysis bath can cause dangerous alterationsof ventricular repolarization and are associated with an increasedincidence of intradialytic cardiac arrests [30–32]. Given the uncer-tainty that exists regarding the underlying mechanisms of suddendeath in dialysis patients, studies performed using long-lastingmonitoring systems could be extremely useful to better understand

the etiology of the phenomenon and to put in place preventivemeasures.

Our study has some limitations such as the retrospective nature, thedifferent follow-up lengths among groups, the inability to perform ICDinterrogation and to go back to the exact time of the onset of the indica-tion for ICD implantation.

In particular, data from the device's interrogation would have had agreat importance and could have made our results more convincing.Unfortunately, given the retrospective nature of the study, these dataare not available.

5. Conclusions

Our results provide some important information about the missingpieces in the scenario of ESRD patients who die of sudden death, firstof all about the size of the problem. To our knowledge these are thefirst data estimating the prevalence of ICD indication in a large popula-tion of dialysis patients. Moreover the study shows howmany ESRD pa-tients with andwithout ICD indication die suddenly. Our study suggeststhat dialysis patients may benefit from ICD if there is an indication forICD implantation, therefore it is not right to deprive ESRD subjectswith a life expectancy greater than one year of this therapeutic opportu-nity. The study also highlights that there is a large proportion of dialysispatients in whom the causes of sudden death could be different fromthose occurring in the population with preserved renal function. Wethink it is important to better understand these differences, becausethey may depend at least partially on modifiable risk factors. Greatattention must be paid to the composition of the dialysis bath, particu-larly in patients at increased risk of sudden death (for example subjectswith diabetes mellitus and coronary disease, also in the presence of apreserved left ventricular ejection fraction). Even the formulation of adialysis prescription that does not involve a long interdialytic intervalmight be helpful. Larger prospective studies and randomized trials areneeded to provide additional elements on how to dealwith the problemof sudden death in ESRD patients.

Conflict of interest

None declared.

Acknowledgments

None declared.

References

[1] S. Genovesi, M.G. Valsecchi, E. Rossi, et al., Sudden death and associated factors in ahistorical cohort of chronic haemodialysis patients, Nephrol. Dial. Transplant. 24(2009) 2529–2536.

177S. Genovesi et al. / International Journal of Cardiology 186 (2015) 170–177

[2] Renal Data System, USRDS 2013 Annual Data Report: Atlas of Chronic Kidney Disease andEnd-Stage Renal Disease in the United States, National Institutes of Health, National Insti-tute of Diabetes and Digestive and Kidney Diseases, Bethesda, MD, 2013.

[3] A.J. Moss, W.J. Hall, D.S. Cannom, et al., Improved survival with an implanted defibrillatorin patients with coronary disease at high risk for ventricular arrhythmia. MulticenterAutomatic Defibrillator Implantation Trial Investigators, N. Engl. J. Med. 335 (1996)1933–1940.

[4] A.J. Moss, W. Zareba,W.J. Hall, et al., Prophylactic implantation of a defibrillator in patientswith myocardial infarction and reduced ejection fraction. Multicenter Automatic DefibrillatorImplantation Trial II Investigators, N. Engl. J. Med. 346 (2002) 877–883.

[5] G.H. Bardy, K.L. Lee, D.B. Mark, et al., Amiodarone or an implantable cardioverter defibrillatorfor congestive heart failure, N. Engl. J. Med. 352 (2005) 225–237.

[6] A.M. Russo, R.F. Stainback, S.R. Bailey, et al., ACCF/HRS/AHA/ASE/HFSA/SCAI/SCCT/SCMR2013 appropriate use criteria for implantable cardioverter-defibrillators and cardiacresynchronization therapy: a report of the American College of Cardiology Foundation ap-propriate use criteria task force, Heart Rhythm Society, American Heart Association,American Society of Echocardiography, Heart Failure Society of America, Society forCardiovascular Angiography and Interventions, Society of Cardiovascular ComputedTomography, and Society for Cardiovascular Magnetic Resonance, Heart Rhythm.10 (2013) e11–e58.

[7] A.Wase, A. Basit, R. Nazir, et al., Impact of chronic kidney disease upon survival among im-plantable cardioverter-defibrillator recipients, J. Interv. Card. Electrophysiol. 11 (2004)199–204.

[8] P.S. Cuculich, J.M. Sànchez, R. Kerzner, et al., Poor prognosis for patients with chronic kid-ney disease despite ICD therapy for the primary prevention of sudden death, Pacing Clin.Electrophysiol. 30 (2007) 207–213.

[9] T. Schefer, T.Wolber, C. Binggeli, J. Holzmeister, C. Brunckhorst, F. Duru, Long-term predic-tors of mortality in ICD patients with non-ischaemic cardiac disease: impact of renal func-tion, Europace 10 (2008) 1052–1059.

[10] M.P. Turakhia, P.D. Varosy, K. Lee, et al., Impact of renal function on survival in patientswith implantable cardioverter-defibrillators, Pacing Clin. Electrophysiol. 30 (2007)377–384.

[11] C.S. Hager, S. Jain, J. Blackwell, B. Culp, J. Song, C.D. Chiles, Effect of renal function on sur-vival after implantable cardioverter defibrillator placement, Am. J. Cardiol. 106 (2010)1297–1300.

[12] F.G. Hage, W. Aljaroudi, H. Aggarwal, et al., Outcomes of patients with chronic kidney dis-ease and implantable cardiac defibrillator: primary versus secondary prevention, Int. J.Cardiol. 165 (2013) 113–116.

[13] P.L. Hess, A.S. Hellkamp, E.D. Peterson, et al., Survival after primary prevention implantablecardioverter-defibrillator placement among patients with chronic kidney disease[publishedonline ahead of print July 18, 2014] Circ. Arrhythm. Electrophysiol. (July 18, 2014)http://dx.doi.org/10.1161/CIRCEP.114.001455.

[14] P.H. Pun, S.M. Al-Khatib, J.Y. Han, et al., Implantable cardioverter-defibrillators for primaryprevention of sudden cardiac death in CKD: a meta-analysis of patient-level data from 3randomized trials, Am. J. Kidney Dis. 64 (2014) 32–39.

[15] A. Aggarwal, Y.Wang, J.S. Rumsfeld, J.P. Curtis, P.A. Heidenreich, Clinical characteristics andin-hospital outcome of patients with end-stage renal disease on dialysis referred for im-plantable cardioverter-defibrillator implantation, Heart Rhythm. 6 (2009) 1565–1571.

[16] N. Makki, P.D. Swaminathan, J. Hanmer, B. Olshansky, Do implantable cardioverter defi-brillators improve survival in patients with chronic kidney disease at high risk of suddencardiac, Europace 16 (2014) 55–62.

[17] M. Cleves, R.G. Gutierrez, W. Gould, Y. Marchenko, An Introduction to Survival AnalysisUsing Stata, Stata Press, Texas, US, 2010.

[18] S. Richard, R.W. Makuch, A non-parametric graphical representation of the relationshipbetween survival and the occurrence of an event: application to responder versus non re-sponder bias, Stat. Med. 3 (1984) 35–44.

[19] N. Mantel, D.P. Byar, Evaluation of response-time data involving transient states: an illus-tration using heart-transplant data, J. Am. Stat. Assoc. 69 (1974) 81–86.

[20] M.K. De Bie, J.C. Lekkerkerker, B. van Dam, et al., Prevention of sudden cardiac death: ratio-nale and design of the Implantable Cardioverter Defibrillators in Dialysis patients trial—aprospective pilot study, Curr. Med. Res. Opin. 24 (2008) 2151–2157.

[21] C.A. Herzog, S. Li, E.D.Weinhandl, J.W. Strief, A.J. Collins, D.T. Gilbertson, Survival of dialysispatients after cardiac arrest and the impact of implantable cardioverter defibrillators,Kidney Int. 68 (2005) 818–825.

[22] S. Hiremath, S.R. Punnam, S.S. Brar, et al., Implantable defibrillators improve survival inend-stage renal disease: results from a multi-center registry, Am. J. Nephrol. 32 (2010)305–310.

[23] F. Khan, E. Adelstein, S. Saba, Implantable cardioverter defibrillators confer survival benefitin patients with renal insufficiency but not in dialysis-dependent patients, J. Interv. Card.Electrophysiol. 28 (2010) 117–123.

[24] P.H. Pun, A.S. Hellkamp, D. Gillian, G.D. Sanders, et al., Primary prevention implantablecardioverter defibrillators in end-stage kidney disease patients on dialysis: a matchedcohort study, Nephrol. Dial. Transplant. 17 (2014)http://dx.doi.org/10.1093/ndt/gfu274(Advance Access published November).

[25] E. Martens, M.F. Sinner, J. Siebermair, et al., Incidence of sudden cardiac death in Germany:results from an emergency medical service registry in Lower Saxony[published onlineahead of print July 24, 2014] Europace (July 24, 2014)http://dx.doi.org/10.1093/europace/euu153.

[26] S. Genovesi, E. Rossi, M. Nava, et al., A case series of chronic haemodialysis patients:mortality, sudden death, and QT interval, Europace 15 (2013) 1025–1033.

[27] R.S. Parekh, L.C. Plantiga, W.H.L. Kao, et al., The association of sudden cardiac death withinflammation and other traditional risk factors, Kidney Int. 74 (2008) 1335–1342.

[28] C. Drechsler, S. Pilz, B. Obermayer-Pietsch, et al., Vitamin D deficiency is associated withsudden cardiac death, combined cardiovascular events, andmortality in haemodialysis pa-tients, Eur. Heart J. 31 (2010) 2253–2261.

[29] A.J. Bleyer, J. Hartman, P.C. Brannon, et al., Characteristics of sudden death in hemodialysispatients, Kidney Int. 69 (2006) 2268–2273.

[30] S. Genovesi, C. Dossi, M.R. Viganò, et al., Electrolyte concentration during haemodialysisand QT interval prolongation in uraemic patients, Europace 10 (2008) 771–777.

[31] P.H. Pun, R.W. Lehrich, E.F. Honeycutt, C.A. Herzog, J.P. Middleton, Modifiable risk factorsassociated with sudden cardiac arrest within hemodialysis clinics, Kidney Int. 79 (2011)218–227.

[32] P.H. Pun, J.R. Horton, J.P. Middleton, Dialysate calcium concentration and the risk of suddencardiac arrest in hemodialysis patients, Clin. J. Am. Soc. Nephrol. 8 (2013) 797–803.