OUP user menu

★ Editor's Choice ★

Sensitivity and positive predictive value of implantable intrathoracic impedance monitoring as a predictor of heart failure hospitalizations: the SENSE-HF trial

Viviane M. Conraads, Luigi Tavazzi, Massimo Santini, Fabrizio Oliva, Bart Gerritse, Cheuk-Man Yu, Martin R. Cowie
DOI: http://dx.doi.org/10.1093/eurheartj/ehr050 2266-2273 First published online: 28 February 2011


Aims Early recognition of impending decompensation and timely intervention may prevent heart failure (HF) hospitalization. We investigated the performance of OptiVol® intrathoracic fluid monitoring for the prediction of HF events in chronic HF patients newly implanted with a device (implantable cardioverter-defibrillator with or without cardiac resynchronization therapy).

Methods and results SENSE-HF was a prospective, multi-centre study that enrolled 501 patients. Phase I (double blinded, 6 months) determined the sensitivity and positive predictive value (PPV) of the OptiVol data in predicting HF hospitalizations. Of 58 adjudicated HF hospitalizations that occurred during the first 6 months in Phase I, 12 were predicted by OptiVol (sensitivity = 20.7%). Sensitivity appeared to be dynamic in nature and at the end of Phase I, had increased to 42.1%. With 253 OptiVol detections, PPV for Phase I was 4.7%. Phase II/III (unblinded, 18 months) determined the PPV of the first OptiVol Patient Alert for detection of worsening HF status with signs and/or symptoms of pulmonary congestion. A total of 233 patients noted such an OptiVol alert and for 210, HF status was evaluated within 30 days. Heart failure status had worsened for 80 patients (PPV = 38.1%).

Conclusions An intrathoracic impedance-derived fluid index had low sensitivity and PPV in the early period after implantation of a device in chronic HF patients. Sensitivity improved within the first 6 months after implant. Further studies are needed to assess the place of this monitoring technology in the clinical management of patients with HF.

  • Bioelectrical impedance monitoring
  • Chronic heart failure (HF)
  • Decompensation
  • Hospitalization
  • Diagnosis

See page 2225 for the editorial comment on this article (doi:10.1093/eurheartj/ehr120)


Despite optimal treatment, chronic heart failure (CHF) is a progressive disease, with the risk of clinical instability and repeated hospital admissions. Accruing evidence shows that episodes of heart failure (HF) decompensation unfavourably affect prognosis.1 Fluid accumulation in the lungs over time (typically weeks) is generally believed to precede episodes of overt decompensation in many patients.2 Education of patients and rapid access to HF clinics assist the early recognition and prevention of episodes of deterioration. In spite of these measures, signs, and symptoms of deterioration are often non-specific or not recognized by the patient, and therefore not acted on in a timely fashion.

Frequent monitoring of patients is of potential interest,36 as this may facilitate identification of patients at risk and allow tailored intervention. The advent of daily remote monitoring has led investigators to test weight change as marker of impending HF decompensation.7,8 Current guidelines indeed recommend monitoring body weight as the method of choice to manage fluid status, but weight gain may not be sensitive enough to detect HF deterioration.35,9 Fluid redistribution, rather than fluid overload or retention10 might explain this diagnostic gap.

The number of CHF patients treated with implantable electrical devices, including cardiac resynchronization (CRT), implantable cardioverter-defibrillator (ICD) therapy, or a combination (CRT-D), is increasing. Current devices incorporate diagnostic features, such as the possibility to monitor thoracic impedance.11Accumulation of intrathoracic fluid decreases the impedance to electrical current passed across the lung. In the Medtronic Impedance Diagnostics in Heart Failure (MIDHeFT) study, impedance measured between the right ventricular lead and the device box correlated strongly with pulmonary capillary wedge pressures in hospitalized patients.12 In a multi-centre observational and non-blinded registry involving 373 CHF patients, intrathoracic impedance monitoring detected clinical HF deterioration with 60% sensitivity and with a positive predictive value (PPV) of 60%.13

This article describes the results of the prospective, multi-centre, double-blinded (Phase I), and open label (Phase II/III) SENSE-HF (Sensitivity of the InSync Sentry OptiVol feature for the prediction of Heart Failure) Study. The OptiVol feature performs a daily measurement of electrical impedance between device box and the right-ventricular electrode, and derives variables for monitoring fluid status. The aims of the present trial were to determine the sensitivity and the PPV of OptiVol fluid monitoring for the detection of HF-related hospitalizations with signs and/or symptoms of pulmonary congestion in CHF patients implanted de novo with an ICD or CRT-D device (Phase I), and the PPV of the first OptiVol Patient Alert for the detection of worsening HF status with signs and/or symptoms of pulmonary congestion (Phase II). The primary objective of Phase III, to describe application of intrathoracic impedance monitoring in current practice, will be addressed separately. However, OptiVol sensitivity is reported for Phase II and Phase III combined.


A complete description of the study design and the endpoint definition has been published previously.14 Briefly, SENSE-HF was a prospective, multi-centre, international study of the OptiVol feature present in the InSync Sentry®, Concerto®, Consulta® (CRT-D), or the Virtuoso® and Secura® (ICD) devices (Medtronic, Inc., Minneapolis, MN, USA). Patients with at least one HF-related hospitalization that required an increase in medication (inotropes, nitrates, diuretics) within the last 12 months were eligible for inclusion. Subject enrolment occurred after successful device implantation and before 34 days post-implant (at which time the OptiVol Fluid Trend Data become available). The study is divided into three phases, illustrated in Figure 1. In Phase I, which started 34 days post-implant, both patient and physician were blinded to the fluid monitoring data during a 6 month period, aiming to determine the sensitivity and PPV of the OptiVol fluid monitoring data in the prediction of HF hospitalizations. In Phase II the physician had access to the fluid monitoring data and the audible patient alert was programmed on. The primary objective of Phase II was to determine the PPV of the first OptiVol Patient Alert with respect to the detection of worsening HF status with signs and/or symptoms of pulmonary congestion. According to the protocol, upon an alert, patients were instructed to consult their treating physician within 7 days for assessment of HF status (‘alert visit’), and, if no signs or symptoms of congestion, a second alert visit was to be scheduled within 12 and 30 days from the alert. During Phase III the physician could optimize the use of intrathoracic impedance data to guide patient management. A blinded Adverse Event Advisory Committee (AEAC) adjudicated all reported hospitalizations from all three phases of the trial for HF relatedness.

Figure 1

Overview of study set up and patient flow. The objectives and main characteristics of the phases are described. PPV, positive predictive value.

The OptiVol feature performs a series of electrical impedance measurements between device box and the defibrillation electrode located in the right ventricle, which are automatically combined to determine an average ‘daily impedance’. Trends in the measured daily impedance determine a calculated ‘reference impedance’. The daily impedance measured for the first 30 days after implantation is excluded to allow for wound healing. The reference impedance is automatically initiated on the 34th day following implant, based on the daily impedances recorded for 4 days prior. Thus, the reference impedance is automatically initiated based on the tacit assumption that the fluid level is normal during this time period. Time periods, during which the measured daily impedance is consistently below the reference impedance, determine the ‘fluid index’.15 The fluid index incorporates both duration and magnitude of the difference between the daily and the reference impedance, and is reset to 0 when the daily impedance is sustained above the reference impedance. When the fluid index crosses a programmable threshold (programmed to 60 Ohm-days nominal in Phases I and II), the device can activate a brief daily audible ‘patient alert’ signal (programmed Off in the blinded Phase I, On in the non-blinded Phases II and III). Figure 2 shows OptiVol data over time of a patient, with daily and reference impedance in the lower part, and index and threshold in the upper part.

Figure 2

Fluid monitoring data is shown from a study patient with an undetected hospitalization near the time of implant. The cardiac resynchronization therapy-defibrillator device generated report consists of a lower graph including the averaged ‘daily impedance’ (black line) and the calculated expected or ‘reference impedance’ (grey line) and an upper graph showing the derived ‘fluid index’ (black line) and programmed fluid index threshold (grey line). The device was implanted in February and the reference impedance and fluid index were established in early March. Superimposed shaded regions identify the dates and durations of two separate heart failure hospitalizations. Note that the daily impedance remained somewhat flat after establishment of the reference impedance until the first hospitalization in May. Hence, the fluid index did not identify this event. However, intravenous diuresis during the hospitalization resulted in a distinct increase in daily impedance and subsequent gradual decrease in impedance following discharge. The daily impedance continued to gradually decline, eventually resulting in a fluid index threshold crossing in early July. The fluid index remained above threshold until the subject was readmitted to the hospital in early August (true positive event). Once again, diuresis resulted in a sharp increase in daily impedance and a subsequent drop in impedance immediately following discharge. (Dashed vertical bars represent date of therapeutic device re-programming.)

This study was conducted in adherence to the Declaration of Helsinki and approved by the Ethics Committees of all participating centres. Informed consent was obtained from all patients. Clinical trial registration: NCT00400985.

Statistical analysis—pre-specified

The endpoint for Phase I was hospitalization (overnight stay) with signs and/or symptoms of pulmonary congestion (as confirmed by the AEAC; events without adequate documentation where excluded as endpoint). Sensitivity and PPV were based on endpoints with date of admission 34 days or more after device implantation and before the 6 months follow-up visit. Excluded were three endpoints that were preceded by surgical revision of the implanted system >14 days after initial implant.

Daily OptiVol measurements were available from stored device data. Threshold crossings were defined as days on which the index was ≥60 Ohm-days, and for which the index was <60 Ohm-days on the day before. Endpoints were classified true positive (TP) when there was a threshold crossing between 1 and 30 days before the day of hospital admission. Sensitivity was calculated as number of TP divided by number of endpoints. Positive predictive value was calculated as the number of TP divided by number of threshold crossings (excluding those <30 days before the 6 months visit). Sensitivity and PPV are reported as percentage. Logistic regression models with generalized estimating equations correction were used to obtain estimates and 95% lower confidence limits (LCLs) corrected for multiple endpoints or threshold crossings per patient.

The study size was determined from the requirement to obtain a 95% LCL for sensitivity within 12% of the point estimate. For true sensitivity = 70%, minimally 54 endpoints were needed, to be obtained from 500 patients followed for 6 months.14

The objective of Phase II was to determine the PPV of the first patient alert with respect to detection of a worsening HF status with signs and/or symptoms of pulmonary congestion, as determined by the attending physician during in-office visits (most often not hospitalizations) within 30 days after the alert. The denominator was the number of patients with any assessment of HF status between 0 and 30 days after the first reported alert. A patient was counted as TP when worsening HF was confirmed.

Statistical analysis—post hoc

Review of Phase I sensitivity results suggested a previously unobserved time effect. Evolution of sensitivity as a function of time since implant was assessed by stratifying endpoints into three equal-sized groups defined by tertiles of time since implant. Sensitivity in the three strata was compared using the Cochran–Mantel–Haenszel test. Sensitivity for detecting adjudicated HF hospitalizations is reported for Phases II/III.

Based on observations from the data, patients with early HF hospitalizations (first and second tertile) were compared with other patients. The early evolution of daily impedance was assessed by means of a repeated measures regression model, with subgroup identifier and time since implant (piecewise linear, separate parameters for the subgroups) as explanatory variables. The average early evolution was compared between patients with and without early HF hospitalizations using contrasts.

Unless indicated otherwise, results are presented as mean ± standard deviation for continuous variables and as number and percentage for categorical variables. A P-value <0.05 is considered significant. All analyses are done in SAS 9.2 (SAS Institute, Cary, NC, USA) by a Medtronic Statistician (B.G.). Authors had full access to the data and analyses. Statistical analyses were reviewed by an independent academic statistician (Prof. Ian Ford).


A total of 501 patients from 41 centres in 13 countries were enrolled between 4 March 2005 and 9 September 2008. The total duration of follow-up in the three phases of the trial was 1.3 ± 0.7 years with a total of 659 patient years. Follow-up in Phases II/III was stopped by the sponsor in August 2008, after the pre-specified 500 patients had been enrolled and the pre-specified number of 54 endpoints was reached. Thus, Phase I was completed as planned, however, for 78 late-enrolled patients, Phases II/III data are not available.

Study population

Baseline characteristics for the total patient cohort and for those patients that entered Phases II/III are shown in Table 1. Patients were treated according to published guidelines, and, at the time of enrolment, the vast majority presented with New York Heart Association (NYHA) functional class II–III and severely reduced left ventricular ejection fraction. The majority of patients had CRT-D devices implanted (78%), with Left Bundle Branch Block in 66% and QRS duration ≥120ms in 85%.

View this table:
Table 1

Patient characteristics

CharacteristicAll patientsWith Phase II/III data
Total recruitment, n (%)501 (100)371 (74)
Demographics, n (%)
 Gender (male)422 (84)315 (85)
 Age (years)65 ± 1065 ± 10
 Body mass index (kg/m2)27 ± 527 ± 5
Therapy type, n (%)
 CRT-D393 (78)297 (80)
 ICD108 (22)74 (20)
Baseline symptoms
 NYHA class at baseline visit, n (%)
  I27 (5)25 (7)
  II201 (40)152 (41)
  III254 (51)183 (49)
  IV18 (4)11 (3)
Cardiovascular disease and medical history, n (%)
 HF aetiology ischaemic289 (58)225 (61)*
 Hypertension232 (46)173 (47)
 Diabetes147 (29)104 (28)
 Atrial Fibrillation188 (38)138 (37)
 Left Bundle Branch Block291 (58)221 (60)
 QRS (ms)147 ± 36149 ± 36*
 QRS width 120 ms or greater, n (%)384 (77)290 (78)
 LVEF (%)26 ± 826 ± 8
 Systolic blood pressure (mmHg)115 ± 18116 ± 18
 Resting heart rate (b.p.m.)72 ± 1272 ± 12
Medication, n (%)
 Beta-blockade427 (85)318 (86)
 ACE-Inhibitor/ARB436 (87)322 (87)
 Diuretics (spironolactone excluded)449 (90)337 (91)*
 Spironolactone254 (51)195 (53)
 Amiodarone131 (26)89 (24)
  • *P < 0.05 for the comparison between patients with and patients without Phase II/III data. Patients that died or discontinued the study during Phase I are included in the cohort without Phase II/III data.


A total of 449 patients completed the 6 months of Phase I. Before the 6 months visit, 30 patients died, 8 patients discontinued the study after system modification, 5 patients withdrew, and 9 patients were lost to follow-up for other reasons. The Phase I follow-up was 0.5 ± 0.1 years, with a total of 242 patient years. Early enrolled patients (n = 371) continued in Phase II/III for a total duration of 1.0 ± 0.5 years. Table 1 includes baseline characteristics for these patients separately. OptiVol data, collected from device memory at hospital visits, were available for 97% of follow-up days for Phase I and 92% for Phase II/III. As pre-defined, post-system modification data from three patients were excluded from endpoint analysis.


During Phase I, 224 unscheduled hospital visits (37 system or procedure related) were reported in 132 patients. Of these, 58 events, occurring in 42 patients, were adjudicated as endpoints (HF-related hospitalization with signs and/or symptoms of pulmonary congestion). Based on fluid index nominal threshold crossing data, 12 of these endpoints were preceded by a fluid index threshold crossing within 30 days prior, leading to an overall sensitivity for Phase I, of 20.7%. Corrected for multiple events per patient, sensitivity was 20.7% with LCL of 13.3%. Threshold crossing occurred on 253 occasions in 188 patients, reflecting a fluid index threshold crossing rate of 1.38 per patient year and resulting in a PPV for HF hospitalizations of 4.7%. Corrected for multiple crossings per patient, PPV was 5.1% (LCL 3.1%).

During the unblinded Phase II, 210 out of 233 patient- or physician-detected alerts were followed within 30 days by at least one in-hospital evaluation of HF status (in total 385 visits, including 268 alert visits, 86 regular follow-up visits, 27 hospitalizations and 4 adverse events). On 80 occasions, HF status had worsened according to the investigator, giving a PPV for HF worsening of 38.1% (LCL 32.8%). In Phases II and III the alert rate was higher than in Phase I (2.96 crossings per patient year).

Post hoc analyses

Sensitivity as a function of time

As illustrated in Figure 3, the rate of admission for HF decompensation was higher in the first months of Phase I, and decreased with time. The proportion of TPs shows the reverse pattern. The consequences of these observed time-related differences in terms of sensitivity are shown in Figure 4, where HF hospitalizations are grouped into tertile groups based on time since implant (first tertile: 34–63 days, second: 64–147 days, third: 148 days or more). Sensitivity was 5.3, 15.0, and 42.1%, respectively (P = 0.016).

Figure 3

This figure shows the number of adjudicated heart failure hospitalizations in time after implant/enrolment for Phase I. Phase I started 34 days after implant and ended 5.8 ± 1.3 months later. The number of hospitalizations is initially higher and decreases with time.

Figure 4

The OptiVol sensitivity significantly increased during Phase I (P = 0.016). Sensitivity is shown for a stratification of Phase I endpoints based on tertiles of time between device implant and hospital admission (cut points 63 days, 147 days).

Phase II/III was unblinded and alerts may have influenced patient and/or investigator actions. The reported 203 unscheduled hospital visits were adjudicated by the AEAC, as in Phase I (blinded). Based on 41 HF hospitalizations, adjudicated by the AEAC, sensitivity was 39.0%.

Sensitivity according to previous and current definitions

Data were analysed post hoc, using the definition of sensitivity reported in the MIDHeFT and Fluid Accumulation Status Trials (FAST).12,16 In these studies, sensitivity was defined as the percentage of events in which the fluid index exceeded a predefined level of 60 Ohm-days at any point in the 30 day window preceding the event without requiring an actual crossing. Thus, HF hospitalizations with fluid index already >60 Ohm-days before the 30 day window are also considered TP. This post hoc analysis results in a sensitivity of 29.3% in blinded Phase I (5.3, 30.0, 52.6% for tertiles of time since implant, respectively, P = 0.006) and 65.9% in unblinded Phase II/III.

Early impedance trends

To better understand the observed low initial sensitivity, a piecewise linear regression model was fitted to the measured daily impedance as function of time since implant. Patients with early HF hospitalization (first 147 days after implant, n = 29) had significantly lower initial impedance values compared with patients without such an event (P = 0.017), and had significantly smaller impedance increase before day 21 (P = 0.026). Likewise, the value at initialization of the reference impedance at day 34 was significantly lower for patients with early events (58.5 ± 9.0 vs. 64.1 ± 8.4 Ohms, P = 0.024). Figure 2 shows an example of OptiVol data for a study patient with an early HF hospitalization and low impedance at initialization of the reference impedance.


SENSE-HF is the first large prospective, double-blinded study (during Phase I) to assess sensitivity and PPV of intrathoracic impedance monitoring in CHF patients, implanted de novo with a device that incorporates this feature. The following observations emerged from this study.

First, overall sensitivity of intrathoracic impedance for the detection of HF hospitalizations with signs and symptoms of pulmonary congestion was 20.7% in Phase I. During the 6 months of Phase I, sensitivity increased significantly from 5.3% in the first 63 days after implant to 42.1% from day 148 onwards, indicating the dynamic nature of sensitivity following implant. Although Phases II/III were unblinded, post hoc analysis showed a sensitivity of 39.0%, which is consistent with the increase seen towards the end of Phase I. Second, for Phase I, the PPV for HF hospitalization with signs and/or symptoms of pulmonary congestion was 4.7%. For Phase II, PPV for signs and/or symptoms of worsening HF after an alert was 37.9%. Third, the event rate was highest early after implantation and decreased thereafter. Impedance trend and initial reference impedance were significantly different in patients with early events compared with patients that did not experience an event in the first 147 days following implant. These observations suggest instability of patients early after implant and might explain lower sensitivity early in Phase I.

Several studies previously assessed the sensitivity of the OptiVol algorithm to predict HF events and have reported results in the range of 60–80%.12,13,1619 Importantly, these studies differed from SENSE-HF in terms of patient population,12 blinding,13,1719 definition of HF events,16,19 analysis of data, or interval between implantation and enrolment.16

In contrast to MIDHeFT12 and FAST,16 the majority of HF hospitalizations in the present study, included in the estimation of sensitivity and PPV (Phase I), occurred in the early period after implantation of the device. During that period, many patients were not clinically stable as evidenced by the high event rate and low initial intrathoracic impedance. By the end of Phase I, the event rate had fallen and surviving patients had improved functional class (NYHA class 2.1 ± 0.7 vs. 2.5 ± 0.7, P < 0.0001). Coinciding with this trend, post hoc analysis of the evolution of sensitivity of the fluid index with time, demonstrated a steady increase to 42.1% by the end of Phase I (Figure 4). Clinical instability, and initialization of the reference impedance at the time of hypervolaemia, are possible explanations for the disappointingly high initial number of endpoints that were not detected by fluid monitoring. This hypothesis is supported by the findings of lower initial impedance and postponed increase in the impedance in patients with early events.

The definition of a TP endpoint used in MIDHeFT and FAST requires the index to be higher than the 60 Ohm-days threshold within 30 days of hospital admission. In SENSE-HF, this criterion was more stringent, being narrowed down to the occurrence of an actual crossing of the threshold within 30 days preceding the event. Post hoc analysis, using the criteria applied in MIDHeFT and FAST, yields higher sensitivity values of 29.3% for Phase I and 65.9% for Phase II/III, which are more consistent with earlier published reports.12,13,1620,23,24

The values for PPV for Phase I and Phase II differed substantially, but are based on different event definitions: in Phase I only adjudicated HF hospitalizations were included but in Phase II in addition to this, worsening HF status with signs and/or symptoms of pulmonary congestion diagnosed by the attending physician at clinic visit was included (only 27 of 385 visits were hospitalizations). When these ‘milder’ HF decompensation events are included, the PPV of OptiVol is substantially increased. The effect of unblinding of patients and physicians may also have played a significant role.

Different approaches, such as monitoring of weight, blood pressure, heart rate, and rhythm, have been tested to predict HF-related hospital admissions.3,5 Information obtained from a totally implantable continuous haemodynamic monitor in the COMPASS-HF trial led to a non-significant 21% lower rate of HF-related events.7 In another study, impedance cardiography correlated moderately with cardiac output but not with left ventricular filling pressures and failed to provide prognostic information.20 Additionally, animal experiments suggest that the use of a left ventricular lead might improve the discriminative power of the impedance vector.21 Additional non-pre-specified analyses demonstrate that the OptiVol measurements of intrathoracic impedance correlate strongly with fluid changes after intravenous diuretic administration (see Supplementary material online).

It is possible that integration of other parameters that are currently recorded by implantable electrical devices, including heart rate variability,22 atrial fibrillation burden, and patient activity,23 will provide a more sensitive tool to predict HF worsening than intrathoracic impedance monitoring alone.6 Also, recent prospective analyses18,24 show that HF patients with fluid index threshold crossings prior to a clinic visit have a significantly higher risk of hospitalization after that visit. Therefore, instead of solely depending on threshold crossing of the fluid index to predict acute decompensation in the short term, intrathoracic impedance monitoring may assist in longer-term prognostication.

Sensitivity later in Phase I increased to a level that is consistent with data reported in other studies,12,13,16 but still unsatisfactory to accurately predict imminent hospitalization. From the present data it seems advisable to clinically confirm euvolemia and appropriateness of the reference impedance at the time fluid monitoring is started. Technical improvements and (ongoing) outcome trials are required to establish the role of intrathoracic impedance monitoring of patients with chronic HF (NCT00480077,25 NCT 00885677,26 NCT00769457).

Study limitations

During Phase I, both the investigator and the patient were blinded to intrathoracic impedance values and the fluid trend data. In contrast, Phases II and III allowed open access to these data by the treating physician and crossing of the fluid index threshold led to an audible alert. Such an alert may have triggered patients to change their medication or fluid or salt intake, potentially leading to a false positive classification of the alert. The alert may also have influenced the physician's assessment of HF status and the data reported to the AEAC. Therefore, the analysis of sensitivity in Phases II and III is less objective, although as in Phase I, each hospital visit or admission was adjudicated by the AEAC in a blinded fashion. Late-enrolled patients have contributed only to Phase I, and comparison of sensitivity between study phases may have been influenced by different characteristics of these patients. However, sensitivity for patients who went on to Phase II/III was not qualitatively different from sensitivity for the total cohort.

According to ruling implant guidelines at the time patients were recruited for enrolment in SENSE-HF, CRT (78%) was overused, since 38% of CRT-D patients were in functional NYHA class II.


The study has been funded by Medtronic Bakken Research Center B.V.

Conflict of interest: The authors received consultancy or research grants from Medtronic and B.G. reports being employed by Medtronic Bakken Research Center, Maastricht, The Netherlands.


We thank Ian Ford for performing an independent review of statistical analyses. We are very grateful to Sandra Jacobs and Doug Hettrick for manuscript support, and the SENSE-HF Study team for study management work on the trial.


CountryPlacePrincipal investigator
BelgiumAntwerpenV. Conraads
BelgiumLiègeJ. Boland
ChinaHong KongC.M. Yu
Czech RepublicPragueJ. Kautzner
Czech RepublicPragueM. Taborsky
DenmarkCopenhagenC. Hassager
FranceGrenobleP. Defaye
FranceLilleS. Kacet
FranceParisF. Pousset
FranceRennesC. Leclercq
FranceSaint DenisG. Lascault
FranceStrasbourgM. Chauvin
FranceRouenF. Anselme
GermanyAachenP. Schauerte
GermanyBad OeynhausenB. Lamp
GermanyBerlinB. Keweloh
GermanyBielefeldD. Meyer zu Vilsendorf
GermanyHeidelbergA. Bauer
GermanyTübingenV. Dörnberger
GermanyMünchenE. Hoffmann
GermanyMünchenS. Schmieder
GreeceAthensT. Maounis
GreeceHeraklion/CreteP. Vardas
GreeceThessalonikiV. Vassilikos
IsraelPetach-TikvaB. Strasberg
IsraelRamat-Gan (Tel Aviv)M. Glikson
ItalyFerraraR. Ferrari
ItalyFirenzeL. Padeletti
ItalyMestreA. Raviele
ItalyMilanF. Oliva
ItalyMilanM. Frigerio
ItalyPaviaM. Landolina
ItalyRomeM. Santini
ItalyTriesteG. Sinagra
PolandWroclawP. Ponikowski
Slovak RepublicBratislavaR. Hatala
Slovak RepublicKosiceJ. Sedlak
Slovak RepublicKosiceB. Stancak
SpainMadridI. Lozano
SpainVigo (Pontevedra)J. Beiras Torrado
UKLondonM. Cowie
UKLondonP. Elliott
UKLondonV. Paul
Appendix 1

List of principal investigators involved in the SENSE-HF Study


Steering Committee
 Prof. Dr V. ConraadsAntwerpen, Belgium
 Prof. Dr M.R. CowieLondon, UK
 Prof. Dr L. TavazziCotignola, Italy
 Prof. C.M. YuHong Kong, China
Adverse Event Advisory/Endpoint Committee
 Prof. S.D. AnkerBerlin, Germany
 Prof. K. DicksteinStavanger, Norway
 Prof. G. FilippatosAthens, Greece
Appendix 2

Study committees’ memberships


View Abstract