OUP user menu

★ Editor's choice ★

Extended heart failure clinic follow-up in low-risk patients: a randomized clinical trial (NorthStar)

Morten Schou, Finn Gustafsson, Lars Videbaek, Chr Tuxen, Niels Keller, Jens Handberg, Anne Sejr Knudsen, Geert Espersen, John Markenvard, Kenneth Egstrup, Hans Ulriksen, Per R. Hildebrandt
DOI: http://dx.doi.org/10.1093/eurheartj/ehs235 432-442 First published online: 8 August 2012


Background Outpatient follow-up in specialized heart failure clinics (HFCs) is recommended by current guidelines and implemented in most European countries, but the optimal duration of HFC programmes has not been established. Nor is it known whether all or only high-risk patients, e.g. identified by NT-proBNP, might benefit from an extended HFC follow-up.

Methods and results In a multi-centre setting, we randomly assigned 921 clinically stable systolic heart failure (HF) outpatients on optimal medical therapy to undergo either an extended follow-up in the HFC (n = 461) or referral back to their general practitioner (GP) (n = 460). The primary composite endpoint was death or a cardiovascular admission. Secondary endpoints included mortality, an HF admission, quality of life, number of days admitted, and number of admissions. The median age of the patients was 69 years; 23% were females; the median left ventricular ejection fraction was 0.30; and the median NT-proBNP was 801 pg/mL; 89% were in NYHA class I–II. The median follow-up was 2.5 years. Time-to-event did not differ between groups (HFC vs. GP) (HR: 1.17, 95% CI: 0.95–1.45, P = 0.149). The two groups did not differ with respect to any of the secondary endpoints at the follow-up (P> 0.05 for all). In high-risk patients identified by NT-proBNP ≥1000 pg/mL, no benefit from HFC follow-up was found (P = 0.721).

Conclusion Irrespective of the level of NT-proBNP stable HF patients on optimal medical therapy do not benefit from long-term follow-up in a specialized HFC in a publicly funded universal access healthcare system. Heart failure patients on optimal medical therapy with mild or moderate symptoms are safely managed by their personal GP.

Trial Registration: www.Centerwatch.com: 173491 (NorthStar).

  • Systolic heart failure
  • Heart failure clinic
  • Randomized clinical trial
  • NT-ProBNP

See page 416 for the editorial comment on this article (doi:10.1093/eurheartj/ehs266)


Despite advances in the treatment and care of heart failure (HF) patients, it is debatable whether the long-term clinical outcome is much improved,1,2 especially after an admission.3 The prevalence of HF is still increasing in many countries4,5 and >14 million individuals suffer from HF in Europe. New strategies for reducing mortality risk and the burden of hospital admissions are urgently needed. As health-care resources are limited development of concepts that can rationalize the use of specialist care is similarly required.

In Denmark guideline authors recommend that all patients with systolic HF are referred to an outpatient HF Clinic (HFC) for up titration of evidence-based therapy such as angiotensin-converting enzyme inhibitors (ACE-I)/angiotensin II receptor blockers (ARB), beta-blockers (BB), aldosterone receptor antagonists (ARA) and evaluation for device therapy as well as disease education, self-management, and physical rehabilitation.6 After optimal medical therapy has been instituted and the disease management programme has been completed, stable patients are referred back to their personal general practitioner (GP). However, the prognosis of these patients is still poor and we hypothesized that high-risk patients might benefit from an extended follow-up in the HFC.

This multi-centre, randomized, controlled trial was conducted to determine whether an extended follow-up in an HFC would reduce the composite endpoint of a cardiovascular (CV) admission or death from any cause among stable systolic HF patients treated with optimal medical therapy. We further hypothesized that a potential effect was driven by the improvement in the outcome of high-risk patients identified by NT-proBNP ≥1000 pg/mL at randomization and, therefore, speculated that the implementation of NT-proBNP could rationalize the use of specialist care.


Study design

The design of the NorthStar study has been published in detail.7 It is an investigator initiated multi-centre, randomized, open-labelled blinded endpoint (PROBE) trial. A steering committee (the members of the steering committee are listed in what follows) designed and oversaw the conduct of the trial and data analysis. The trial progress was monitored by an executive steering committee. Data were collected, managed, and analysed by the investigators according to a predefined statistical analysis plan. The manuscript was prepared by the executive steering committee, whose members had unrestricted access to the data, and was subsequently revised by all authors. All authors made the decision to submit the manuscript for publication and assume responsibility for the accuracy and completeness of the data and analyses. Representative from the study sponsor (Roche Diagnostics) had no influence on the study design, data analysis, or manuscript preparation.

Study patients

The trial was approved by The Danish Ethics Committee (KF 01/2724936). All patients provided written informed consent and the study was performed according to The Helsinki Declaration II. Patients were recruited from 18 out of 40 public HFCs in Denmark from November 2005 to December 2009.

Eligibility criteria were as follows: age of at least 18 years, a left ventricular ejection fraction (LVEF) of no >45% at the baseline visit in the HFC, educated in HF (=educated in HF and self-management) and on optimal medical therapy with an ACE-I/ARB (unless contraindicated), a BB (unless contraindicated) at the recommended—or the maximal tolerated dose, an ARA (unless contraindicated), and an ICD and/or CRT, if indicated. In addition, patients were required to fulfil predefined criteria for clinical stability (=no changes in doses of diuretics, HF symptoms and NYHA class constant, and physical examination without signs of volume overload) at two consecutive visits in the HFC.7

Exclusion criteria were: plasma creatinine >200 µmol/L, waiting for a heart transplant, valvular, or ischaemic heart disease with planned surgery or percutaneus intervention, withdrawal of ACE/ARBs, BB, and ARAs due to a reversible cause of cardiomyopathy, malignancy with life expectancy <5 years, and dementia.

Study procedures

Balanced and simple randomization with strata was used. Patients were randomly assigned by sealed envelopes at every participating site in a 1:1 ratio to either follow-up by the extended HFC programme or usual care by a GP. Allocation was stratified according to NT-proBNP ±1000 pg/mL, a value close to the median of NT-proBNP levels in different HF cohorts.7 In that way it was expected that the present population could be split into a low- and high-risk group. The 18 HFCs were divided into five blocks based on expected included number of patients (balanced randomization). The investigators and the GPs were aware of NT-proBNP levels. In patients randomized to usual care, a discharge letter including the objectives of the study, NT-proBNP value, and medical status was sent to the GPs. The GPs were not trained in relation to the study.

Patients allocated to usual care arranged an individual follow-up programme with their GP, whereas patients allocated to an extended follow-up in the HFC completed the following programme: visits at 1–3-month intervals in the HFC at the discretion of the investigators. Medical treatment was controlled at each visit to ensure adherence, and if adherence decreased education was repeated. Symptoms, weight, blood pressure, and electrolytes were monitored in the clinics, and if the HF disease progressed doses of diuretics were adjusted and potential need for the escalation of guideline therapy was re-evaluated. Comorbidity was also managed in the clinics. The patients had access to free daily telephone consultations with the HF nurse who was supervised by cardiologists. At the end of the follow-up, all living patients from both groups were invited to an end-of-study visit in the HFC. Data were captured in an electronic Case Report Form used in all the HFCs.

Study outcomes

The primary endpoint, which was analysed as the time from randomization to the first event, was a composite of death from any cause or admission for a protocol-specified CV cause. Reasons for such admissions included worsening HF, myocardial infarction, stroke, cardiac syncope, hypertension, unstable angina, and arrhythmia. The secondary endpoints were death from any cause, admission for a CV cause, admission for worsening HF, over all admission, patients admitted, admission days and number of admissions, Minnesota Living with Heart Failure Questionnaire score ( = health-related quality of life), functional class, and plasma level of NT-proBNP at end-of-study visit. The vital status and hospital admissions were evaluated by an independent endpoint committee unaware of study-group assignments and NT-proBNP levels. The committee used ICD codes (criteria) from the validated Danish Personal Registry.8,9 Adjudication of endpoints using direct source data was not performed.

Statistical analysis

Based on data from a meta-analysis,10 it was assumed that the HFC intervention would result in a 25% reduction in the primary composite endpoint. The event rate for the primary composite endpoint in the control group was assumed to be 15% per year, derived from rates in an observational study with similar patients. Assuming a two-sided α of 5% and β of 80%, this led to a sample size of 1000 patients, evenly distributed between the two treatment groups. No interim analyses were planned and randomization was planned to happen from November 2005 to December 2009.7 It was planned that patients should be followed from 6 to 4.5 years (median 2.5 years). Enrollment was stopped at the predefined date, where 921 patients were randomized.

Data from all patients who underwent randomization were analysed according to the intention-to-treat principle. Comparability of baseline characteristics between the two study groups was assessed by an unpaired t-test (parametric) or a Mann Whitney U test (non-parametric) for continuous variables, or χ2 tests for discrete variables. The analyses of the primary composite endpoint and secondary endpoints were performed with the use of Kaplan–Meier estimates, with the log-rank test for the comparison of the study groups, and the Cox proportional hazards model to calculate hazard ratios (HRs) and 95% CI. We compared the number of admissions and admission days, differences in Minnesota living with Heart Failure Score and differences in NT-proBNP by a Mann Whitney U test (non-parametric) and the number of patients admitted by a χ2 test. A subgroup analysis included the assessment of the primary composite endpoint within subgroups based on NT-proBNP ±1000 pg/mL (test for heterogeneity). Other subgroups [±median of age, ±NYHA class III, ±LVEF = 0.30, ±atrial fibrillation, ±estimated glomerular filtration rate = 60 mL/min/1.73 m2, ±hyponatraemia (<136 mmol/L), ±anaemia (<7.4 mmol/L for females and <8.1 mmol/L for males), ±diabetes, ±previous myocardial infarction, ±80 mg of loop diuretics/24 h, ±ARA, and ±25 points on Minnesota Living with Heart Failure Questionnaire, ±per-site enrollment of 100 patients] were also evaluated by the test for interaction between each pair of subgroups and the main treatment effect. Safety data at the follow-up end were compared in a manner similar to the baseline characteristics. Statistical analyses were performed on SPSS version 18.0.



A total of 6180 patients had at least one visit in one of the HFC during the randomization period (Figure 1). Of these, 1640 fulfilled the criteria for stability and were eligible. n = 54 were excluded due to the presence of exclusion criteria [aortic stenosis (n = 4), p-creatinine >200 µmol/L (n = 8), cancer (n = 9), dementia, or communication problems, (n = 20) withdrawal of ACE-I/ARB and BB (n = 5), heart transplant listing n = 5, frequent and severe angina (n = 2), and waiting for aneurism surgery (n = 1)]. Furthermore, 103 declined to participate in a scientific project, 107 wanted to be followed by their GP, 256 patients had a NT-proBNP <1000 pg/mL after allocation in this group was completed and 199 consenting patients were randomly assigned for a parallel study.7 The remaining 921 patients were enrolled; 461 were randomly assigned to undergo follow-up in the HFC and 460 to receive usual care. A total of 723 (91%) of the surviving study patients completed the end-of-study visit, with no between group difference in the rate of completion. Admissions and vital-status verification were completed for 100% of patients. Baseline characteristics of the patients were similar between the two groups (Table 1). Patients were elderly, mostly male and comorbidity was frequently present. All patients were well treated with evidence-based therapy as expected based on the inclusion criteria. On average the patients had been followed already for 9 (2–63) months in the HFC at the time of randomization.

View this table:
Table 1

Baseline characteristics of the patients [median (95% CI) or number (%)]

CharacteristicsHeart failure clinic (n = 460)Usual care (n = 460)
 Age (years, median)69 (47–86)69 (43–86)
 Female (sex)106 (23)122 (27)
 NYHA class I–II411 (89)410 (88)
 Blood pressure (mmHg)
  Systolic127 (90–177)124 (90–166)
  Diastolic75 (50–92)73 (50–100)
 Heart rate (b.p.m.)65 (49–92)66 (48–95)
 Left ventricular ejection fraction0.32 (0.15–0.45)0.30 (0.15–0.45)
 Atrial fibrillation152 (33)146 (32)
 LBBB in non-paced ECG76 (19)92 (23)
 Body mass index (kg/m2)26 (20–37)26 (19–40)
 Heart failure aetiology
  Non-ischaemic188 (41)193 (43)
  Ischaemic268 (59)255 (57)
 Minnesota Living with Heart Failure Questionnaire Score25 (0–75)22 (0–73)
 Duration of the basic HFC programme, months9 (2–61)9 (2–63)
Medical history
 Admission within 12 months188 (41)207 (45)
 Hypertension193 (43)183 (40)
 Myocardial infarction238 (52)219 (48)
 Previous PCI/CABG190 (41)188 (41)
 Stroke or TCI53 (12)56 (12)
 Peripheral vascular disease37 (8)42 (9)
 Stable angina pectoris36 (8)41 (9)
 Diabetes85 (18)85 (18)
 Chronic obstructive pulmonary disease62 (13)73 (16)
Laboratory tests
 NT-proBNP (pg/mL)793 (63–6720)803 (58–7517)
 Haemoglobin (g/dL)8.6 (6.6–10.3)8.5 (6.70–10.4)
 Anaemiaa94 (21)102 (23)
 Sodium (mmol/L)140 (131–145)140 (130–145)
 Hyponatraemiab76 (17)72 (16)
 Potassium (mmol/L)4.3 (3.5–5.1)4.3 (3.5–5.2)
 Creatinine (µmol/L)88 (62–185)91 (62–193)
 Estimated GFRMDRD – mL/min/1.73 m269 (30–118)66 (30–121)
 Estimated GFRMDRD ≤ 60 mL/min/1.73 m2152 (33)179 (39)
 ACE-I/ARB392 (85)408 (89)
 ACE-I/ARB, max. target dose306 (66)313 (68)
 BB388 (84)391 (85)
 BB, at target dose243 (53)226 (49)
 ARA143 (31)153 (33)
 Loop diuretics259 (56)272 (59)
 Furosemide doses (mg/24 h)40 (0–240)40 (0–290)
 Thiazide42 (9)30 (7)
 Digoxin71 (15)66 (14)
 Lipid-lowering agents296 (64)303 (66)
 Antiplatelets322 (70)338 (73)
 Anticoagulants142 (31)117 (25)
 Nitrates29 (6)40 (9)
 Gout medication35 (8)26 (6)
 Antidepressants32 (7)26 (6)
 Amiodarone14 (3)17 (4)
Device therapy
 ICD42 (9)32 (7)
 CRT7 (2)7 (2)
  • NYHA class, New York Heart Association; LBBB, left bundle branch block; HFC, heart failure clinic; PCI, percutaneous intervention; CABG, coronary artery bypass graft; TCI, transitory cerebral ischaemia; NT-proBNP, amino-terminal-pro-brain-natriuretic-peptide; estimated GFRMDRD, estimated glomerular filtration rate (modification of diet in renal disease formula); ACE-I, angiontensin-converting enzyme inhibitors; ARB, angiotensin receptor blocker; BB, beta-blocker; ARA, aldosterone receptor antagonist; ICD, implantable cardiac defibrillator; CRT, cardiac resynchronization therapy.

  • aAnaemia defined according to WHO: <7.4 mmol/L for females and <8.1 mmol/L for males.

  • bHyponatraemia defined as plasma sodium <136 mmol/L.

Figure 1

Screening, randomization, and follow-up of the study patients (trial structure).

Primary endpoint

No significant difference was observed between the two groups in the rate of the primary endpoint, which occurred in 177 patients (200/1000 person-years) in the group followed in the HFC and in 159 patients (172/1000 person-years) in the usual care group. The HR for the primary endpoint of CV admission or death from any cause with the HFC follow-up vs. usual care was 1.17 (95% CI: 0.95–1.45, P = 0.149). Kaplan–Meier time-to-event curves for the composite endpoint of a CV admission or death from any cause did not reveal a significant difference between the two groups (LogRank: P = 0.148) (Figure 2A). NT-proBNP ≥1000 pg/mL identified a high-risk group, but the HFC intervention did not improve the outcome for these patients (HR: 1.11, 95% CI: 0.81–1.50, P = 0.526 for patients with NT-proBNP ≥1000 pg/mL and HR: 1.22, 95% CI: 0.90–1.64, P = 0.447 for patients with NT-proBNP <1000 pg/mL) and any effect modification by NT-proBNP strata was not observed (P = 0.721 for heterogeneity). Kaplan–Meier time-to-event curves for the composite endpoint of a CV admission or death from any cause did not reveal significant differences between the subgroups (Figure 2B) (LogRank: P = 0.196 for NT-proBNP <1000 pg/mL and P = 0.526 for NT-proBNP ≥1000 pg/mL, respectively).

Figure 2

(A) Kaplan–Meier curves for the primary outcome. The primary outcome of death from any cause or hospitalization for cardiovascular causes is shown for patients followed in heart failure clinics and those followed by the general practitioners (usual care). The Kaplan–Meier curves illustrate the time to the first event (hazard ratio in the HFC group, 1.17 (95% CI: 0.95–1.45, P = 0.149; P = 0.148 by the log-rank test). GP, general practice (usual care); HFC, heart failure clinic. (B) Kaplan–Meier curves for the primary outcome stratified according to NT-proBNP ≥1000 pg mL. The primary outcome of death from any cause or hospitalization for cardiovascular causes is shown for patients followed in heart failure clinics and those followed by the general practitioners (usual care) stratified according to NT-proBNP ≥1000 pg/mL. The Kaplan–Meier curves illustrate the time to the first event (hazard ratio in the HFC groupNT-proBNP ≥1000 pg/mL: 1.11 (95% CI: 0.81–1.50, P = 0.526; P = 0.526 by the log-rank test and hazard ratio in the HFC groupNT-proBNP <1000 pg/mL: 1.22 (95% CI: 0.90–1.64, P = 0.447; P = 0.196 by the log-rank test). (P = 0.721 for test heterogeneity). GP, general practice; HFC, heart failure clinic.

Secondary endpoints

No significant differences were seen between the two groups with respect to the secondary endpoints (Table 2). Admission for any cause occurred in 255 patients (55%) in the HFC group and 236 patients (51%) in the usual care group. The HR for admission for any cause with HFC vs. usual care was 1.11 (95% CI: 0.93–1.33, P = 0.242) (LogRank: P = 0.242). A total of 60 patients (13%) in the HFC follow-up group and 64 patients (14%) in the usual care group died during the follow-up period. The HR for death was 1.05 (95% CI: 0.74–1.50, P = 0.777) (LogRank: P = 0.777). Cardiovascular admission (LogRank: P = 0.175), HF admission (LogRank: P = 0.862), the number of days spent in hospital, and the number of admissions were also similar in the two groups (LogRank and Table 2).

View this table:
Table 2

Secondary endpoints

Heart failure clinic, no. of patients with event (n = 460)Usual care, no. of patients with event (n = 460)Hazard ratio 95% CIP-value
Death60641.05 (0.74–1.50)0.777
CV admission1411241.18 (0.93–1.50)0.176
HF admission69671.03 (0.74–1.44)0.862
Over-all admission2552361.11 (0.93–1.33)0.242
Admission days, total438142700.453
Number of admissions, total6556940.378
Patients admitted, total (%)255 (55.3)236 (51.3)0.223
Change in NT-proBNP
Median (pg/mL)1811320.736
Inter-quartile range50–31230–235
Change in Minnesota Living with Heart Failure Questionnaire
Inter-quartile range−1–0−1–1
NYHA class at f/u, total (%)
 I72 (19)69 (20)0.591
 II243 (66)219 (62)
 III57 (15)62 (18)
 IV0 (0)1 (0)
  • CV, cardiovascular; HF, heart failure; NT-proBNP, amino-terminal-pro-brain-natriuretic-peptide; NYHA, New York Heart Association.

Subgroup analyses

Subgroup analyses (Table 3) showed that none of the baseline characteristics of the patients identified a subgroup in which the HFC programme was effective.

View this table:
Table 3

Occurrence of the primary endpoint in subgroups

SubgroupHeart failure clinic, no. of events/no. of patients (%)Usual care, no. of events/no. of patients (%)Hazard ratio (95% CI)P-value for interaction
 NT-proBNP ≥1000 pg/mL85/208 (41)78/203 (38)1.11 (0.81–1.50)0.721
 NT-proBNP <1000 pg/mL92/252 (36)81/257 (32)1.22 (0.90–1.64)
 Age ≤6985/246 (34)73/242 (30)1.18 (0.86–1.61)0.968
 Age >6992/214 (43)86/218 (40)1.18 (0.88–1.58)
 LVEF ≤3096/226 (43)93/245 (38)1.19 (0.89–1.58)0.960
 LVEF >3078/229 (34)65/203 (32)1.13 (0.81–1.57)
NYHA class
 NYHA class I–II149/410 (36)132/413 (32)1.20 (0.95–1.52)0.433
 NYHA class III28/50 (51)27/47 (57)0.96 (0.57–1.63)
Atrial fibrillation
 Yes55/152 (36)44/146 (30)1.23 (0.83–1.85)0.670
 No122/303 (40)115/314 (37)1.14 (0.88–1.47)
Renal function
 eGFR ≤60 mL/min/1.73 m267/152 (44)67/179 (37)1.32 (0.94–1.85)0.431
 eGFR >60 mL/min/1.73 m2110/308 (36)92/281 (33)1.12 (0.85–1.48)
 Yes42/94 (45)33/102 (32)1.50 (0.95–2.36)0.680
 No123/348 (35)128/350 (37)1.08 (0.84–1.37)
 Yes42/76 (55)31/72 (43)1.48 (0.93–2.35)0.253
 No135/384 (35)128/388 (33)1.10 (0.87–1.41)
Minnesota Living with Heart Failure Questionnaire
 Score ≥2598/223 (43)75/192 (39)1.11 (0.82–1.50)0.787
 Score <2566/209 (32)68/233 (29)1.20 (0.85–1.68)
Previous myocardial infarction
 Yes96/238 (40)81/219 (37)1.17 (0.87–1.57)0.279
 No81/222 (36)78/241 (32)1.16 (0.85–1.58)
 Yes39/85 (46)34/84 (41)1.08 (0.68–1.72)0.883
 No53/375 (37)124/375 (33)1.18 (0.93–1.51)
Use of aldosterone antagonist
 Yes59/143 (48)64/153 (42)1.08 (0.75–1.52)0.489
 No118/317 (37)95/307 (31)1.25 (0.96–1.64)
Use of ≥80 mg furosomide/24 h
 Yes70/158 (44)68/151 (45)1.04 (0.74–1.45)0.441
 No107/302 (35)91/309 (29)1.25 (0.95–1.66)
Size of the HFC
 ≥100 patients enrolled87/218 (40)92/224 (41)1.02 (0.76–1.36)0.144
 <100 patients enrolled90/242 (37)67/236 (28)1.39 (1.01–1.91)
  • NT-proBNP, amino-terminal-pro-brain-natriuretic-peptide; LVEF, left ventricular ejection fraction; NYHA, New York Heart Association; eGFRMDRD, estimated glomerular filtration rate (modification of diet in renal disease formula); HFC, heart failure clinic.

  • aAnaemia defined according to WHO: <7.4 mmol/L for females and <8.1 mmol/L for males.

  • bHyponatraemia defined as plasma sodium <136 mmol/L.

Changes in medication and implantation of devices after randomization

Patient reported adherence did not differ between the two groups (Table 4). The patients in the HFCs more frequently had ARAs initiated (P < 0.001) and doses of diuretics changed (P < 0.001). Frequency of ICD and CRT implantation did not differ between the two groups.

View this table:
Table 4

Changes in drug and device therapy during the follow-up

Heart failure clinic (n = 371), no. of patients (%)Usual care (n = 352), no. of patients (%)P-value
ACE-I/ARB initiated23 (6)13 (4)0.126
ACE-I/ARB withdrawn9 (2)14 (4)0.233
BB initiated25 (7)21 (6)0.685
BB withdrawn10 (3)9 (3)0.917
ARA initiated39 (10)13 (4)<0.001
ARA withdrawn23 (6)24 (7)0.721
Loop diuretics initiated48 (13)17 (5)<0.001
Loop diuretics withdrawn26 (7)24 (7)0.936
Loop diuretics dose increased90 (25)49 (14)<0.001
Doses of diuretics at the follow-up end (mg furosomide/24 h), median (95% CI)40 (0–247)30 (0–248)0.037
Tiazide diuretics initiated28 (8)11 (3)0.009
Tiazide diuretics withdrawn15 (4)5 (1)0.033
Digoxin initiated21 (6)14 (4)0.300
Digoxin withdrawn10 (3)8 (2)0.724
ICD implanted15 (4)12 (3)0.664
CRT implanted11 (3)7 (2)0.406
  • ACE-I, angiontensin-converting enzyme inhibitors; ARB, angiotensin receptor blocker; BB, beta-blocker; ARA, aldosterone receptor antagonist; ICD, implantable cardiac defibrillator; CRT, cardiac resynchronization therapy.

Follow-up programme in the general practitioner group and cross-over

At the end-of-study visit, patients allocated to usual care were asked: ‘how often did you visit your GP concerning your heart disease?’ To this 78 (20%) answered ‘each month’, 47 (12%) answered ‘every two months’, 113 (29%) answered ‘every 1–3 months’, 17 (4%) answered ‘every 6 months’, 24 (6%) answered ‘annually’, and 117 (29%) answered ‘occasionally’. Twenty-four patients (5%) of the patients allocated to usual care were followed in the HFC at the end of the follow-up, 13 (3%) of the patients allocated to the HFC were followed by their GP at the follow-up end, 41 (9%) of the patients allocated to usual care were referred back to the HFC, but discharged again and 59 (12%) of the patients allocated to usual care had visited a cardiologist.

Safety data

Safety data at the follow-up end visit did not differ (Table 5). No adverse events were reported during the study.

View this table:
Table 5

Safety data at the follow-up

Heart failure clinic (n = 372), no. of patients (%)Usual care (n = 351), no. of patients (%)P-value
Hypotension (systolic blood pressure <90 mmHg)3 (1)2 (1)0.569
Bradycardia (heart rate <50 b.p.m.)16 (5)16 (5)0.764
Tachycardia (heart rate >100 b.p.m.)7 (2)9 (3)0.488
Hyperkalaemia (p-potassium >5.0 mmol/L)13 (4)22 (7)0.063
Hypokalaemia (p-potassium <3.2 mmol/L)2 (1)0 (0)0.177
Severe hyponatraemia (p-sodium <125 mmol/L)1 (0)1 (0)0.943
>50% increase in p-creatinine13 (4)13 (4)0.798
NYHA class IV1 (0)0 (0)0.591
Death outside the hospital31 (7)27 (6)0.593
  • NYHA, New York Heart Association.


In this multi-centre, randomized, controlled trial involving systolic HF patients on optimal medical therapy, we found no reduction in the risk of CV admission or death associated with an extended follow-up in an HFC when compared with usual care by the GP. Moreover, there was no reduction in the occurrence of any secondary endpoints. In a pre-specified subgroup analysis, NT-proBNP ≥1000 pg/mL identified a high-risk group, but the HFC intervention did not improve the outcome even for these patients.

The design of our trial differ from previous trials evaluating the long-term effect of an HFC intervention since we tested the effect of an HFC clinic intervention in patients on optimal medical therapy who had already completed the basic educational programme ( = education in HF and self-care management). Our results are partly in contrast to the previously published Australian and Brazilian (REMADHE) single-centre studies.11,12 There are, however, significant differences between these studies and the NorthStar study. The results of the Australian study represent a pooled analysis of home-based interventions in somewhat older and predominantly class III–IV patients, and in the Brazilian study adherence to the follow-up programme was 45% in contrast to 97% observed in our study. Furthermore, 91% of the patients visited the HFCs at the follow-up end and only 8% crossed over in our study. In the GESICA trial (extension),13 patients were on optimal therapy before randomization, and the investigators observed an effect on HF admission after 3 years of the follow-up, but all-cause admission was not reported. Moreover, patients included in the GESICA trial were more symptomatic and required more diuretics than the patients in our study. Furthermore, in GESICA the patients received BB treatment less frequently, which could impact the outcome significantly. Likely, the patients in the NorthStar study were clinically more stable and had completed the basic educational programme, which may explain the difference in the outcome between the studies.

Together with the COACH,14 HART,15 TELE-HF,16 and TIM-HF17 trials our data challenge the results of recent meta-analyses10,18 suggesting a long-term effect of different HFC programmes beyond optimal medical therapy and basic patient HF education. However, extrapolation of the results of the current study to any local health-care system would require that the organization and quality of the primary care system be comparable with the usual care arm of the NorthStar study.19 The primary care system in Denmark is publicly funded with universal access, and >90% of all Danes are treated by a personal GP (family physician). Patients do not pay for consultations or testing at their GP nor do they pay for visits to the emergency department. Outpatients do pay for their medications, but only up to ∼500 EUR per year beyond which expenses are reimbursed by the state. It is noteworthy that large multi-centre trials testing different follow-up programmes have been neutral, whereas primarily small single-centre trials have demonstrated an effect of disease management programme intervention. Subgroup analyses from COACH20 and HART15 have also questioned the ‘one size fits all’ educational programme and in the IMPROVE HF Registry HFCs were superior only with regard to CRT implantation and HF education, but doses of ACE-/ARBs and BBs were not reported.21,22 Whether experience from dedicated HFCs can be extrapolated to all HFCs in The Western World and whether there is an effect of different HFC programmes beyond optimal medical therapy and basic education is, therefore, debatable, but it should be kept in mind that the HFC intervention is complex and components are difficult to separate.19,23

The reasons for the lack of benefit are uncertain, but several explanations warrant consideration. We hypothesized that adherence would decrease when patients were referred from the HFC in the hospital setting back to their GP in the primary care setting. This was not the case as estimated by patient reported adherence at the follow-up end (Table 4). We further hypothesized that the GPs threshold for admitting the patients would be lower than in HFCs, which was also disproved. Long-term adherence was good and the number of admissions acceptable in the HFC programme, so the HFC intervention did not fail, but the results of the follow-up by the GPs surpassed our expectations with a low proportion of cross-over. A discharge letter to the GPs including the medical status and the objectives of the study may explain this observation and our results underscore the importance of good communication between the HFC and primary care.

It may be argued that we should have implemented a more intensive HFC follow-up program. However, telemonitoring/structured telephone interviews have not improved the outcome in recently completed multi-centre trials16,17 and in Denmark elderly patients are eligible to receive home visits by a geriatric nurse as part of routine care. Patient reported adherence to drug therapy was >90% and a pharmacist could probably not improve this.24 Whether the implementation of biomarkers can improve the outcome in systolic HF patients on optimal medical therapy is unclear25 and similarly, the efficacy of advanced implantable monitoring systems is yet generally unproven in NYHA class II patients.26,27 However, it is clear that new concepts which can improve the outcome beyond careful clinical care in patients who are considered stable need to be identified in the future, as >50% of the patients in the current study required hospital-admission within 2.5 years.

Selection bias at enrollment is always a serious concern when trial results are reported. At randomization our patients had the same age, LVEF, and prevalence of ischaemic heart disease as the whole Danish Heart Failure Clinic Registry,28 but NYHA class and medication were different which was expected since we intended to investigate clinically stable patients. In theory, patients can be too healthy or too sick to respond to an intervention.29 The NorthStar population had a three-fold increased risk for mortality and CV admissions compared with an age-matched population,30 and a lower event rate than the whole Danish Heart Failure Clinic Registry,28 so we believe that our patients were appropriate for the applied intervention and that selection bias at enrollment cannot explain the lack of effect.

We choose a composite endpoint consisting of mortality and a CV admission. It may be argued that we should have focused on HF admission. However, in a long-term study HFCs may prevent death18 and patients may be admitted due to, e.g. myocardial ischaemia and atrial fibrillation, and not only for decompensation. The mortality rate, CV admission rate, HF admission rate, overall admission rate, number of admissions, and admission days were not improved by the HFC follow-up programme (Table 2). Therefore, the HFC programme had no effect on any secondary endpoint leaving selection of a wrong endpoint as an unlikely explanation of observed neutral result.

We also tested the hypothesis that the implementation of NT-proBNP could rationalize the use of specialist care. NT-proBNP identified high-risk patients, but the intervention could not improve the long-term clinical outcome despite the use of a biomarker for risk stratification.31 In a recently published study of eplerenone in moderate HF, BNP/NT-proBNP was successfully used as an inclusion criterion32 and in a study of a statin in ischaemic cardiomyopathy patients with the lower NT-proBNP levels benefited from statin treatment in a post hoc analysis,33 but our data suggest that NT-proBNP cannot be used to identify patients on optimal medical therapy who can benefit from specialist care for a longer period of time. This is in contrast to data reported for newly discharged patients.34 The latter study did, however, randomize the patients before target doses were reached and it may be argued that all newly discharged HF patients not on optimal therapy should be considered as high-risk patients28 who need follow-up in an HFC.35 All other subgroup analyses were also neutral (Table 3).

Some methodological issues and limitations require consideration. When we designed the study it was routine to refer stable patients on optimal medical therapy back to their GP. In consequence we designed the study as a superiority trial and not as a non-inferiority trial. We may conclude that an extended HFC intervention is not superior to the follow-up with the GP and that patients can be safely referred back, but we have not shown that the two models for the follow-up are equal. We powered the trial to detect a difference of 25% and we may, therefore, have overlooked an effect of 10–15%, which could be of potential clinical relevance.36,37 The NT-proBNP stratified hypothesis was a pre-specified interaction analysis and all unmeasured confounders were balanced due to the randomization process, but it may be argued that the trial was not powered to evaluate subgroups.38 However, given the actual result of the study we believe that a type II error did not likely occur. The generalizability of our results is limited to patients with HF due to systolic dysfunction in functional class I–II. It may be considered that the selection of patients with systolic HF is a strength of the study given the lack of homogeneity of HF with preserved ejection fraction patients and the lack of evidence-based interventions in this patient group. Only 10% of our patients had a device implanted, which is a lower rate for this period than in both the USA21 and Europe.39 Finally, whether the inherent open design of our trial has biased the results in either direction remains unknown.

Irrespective of the level of NT-proBNP stable HF patients on optimal medical therapy do not benefit from long-term follow-up in a specialized HFC in a publicly funded universal access health-care system. Patients on optimal medical therapy with mild to moderate symptoms can safely be discharged to their personal GP implying that the HFC should focus on up titration of medication and basic education of all patients and reserve an extended follow-up for patients with severe symptoms.


The study is funded by unrestricted grants from Roche Diagnostics, Basel, Schwitzerland. Development of electronical case report form is supported by Merck, Sharp and Dohme, Copenhagen, Denmark. The Copenhagen Hospital Corporation supported the study with a research grant to M.S.

Conflict of interest: P.R.H. reports receiving research grants, and consulting and lecture fees from Roche Diagnostics; F.G. reports receiving consulting fees from Roche Diagnostics. M.S. reports receiving lecture fees from Roche Diagnostics. The rest of the authors have no conflicts of interest to report.


NorthStar Trial Committees:

Executive Committee: M.S. (principal investigator), P.R.H. (chair), F.G. (voting), L.V. (voting).

Steering Committee: A.S.-K., C.T., J.H., N.K., J.A., Soren Lind-Rasmussen, Jens Toft, Olav W. Nielsen, Henrik Ryde, K.E., J.M., Ole Nyvad, Helge Andreassen, Jens Herman Jensen, H.U., G.E., Per Sorensen.

Data monitoring nurse: Per H. Nielsen.

Endpoint Validation Committee: Jordi Dahl and Michael Kjaer-Poulsen.

NorthStar Trial Investigators, Aalborg: Eva Korup, Helle Broberg, Vivi Nielsen, Mette Storgaard; Bispebjerg: Henriette P. Pedersen; Esbjerg: Anette Dam, Tove Christensen, Lilli Thude; Fredericia: Thomas Gohr, Bodil S. Madsen, Hanne-Lis Nielsen; Frederiksberg: Jens Rosenberg, Michael Egstrup, Anna-Marie Jensen, Jeanett Hemmingshoj, Hanne Bartholdy, Susanne Groth, Gitte Leth; Frederikshavn: Marianne Jensen, Holger Sejrsen; Glostrup: Ellen Skaarstad, Ingelise Larsen, Jannie Jensen; Herlev: Anne-Mette Borrild, Birthe Mogelhoj, Jimi Claussen, Vibeke Jacobsen; Hillerod: Peter Kaiser-Nielsen, Lone Woelders, Nina Viem, Louise Oechsler; Hvidovre: Inge-Lise Knoefler, Kristina Karlsson; Kolding: Helle Egeberg, Mette R Pedersen; Odense: Janne Milton, Charlotte Anker, Camilla Thomassen, Karin Koertz, Jane Petersen; Roskilde: Else Hebsgaard, Marianne Baeksbo, Lene Frederiksen; Silkeborg: Hanne Skoffer, Camilla Jespersen; Skive: Jette Lynggaard; Slagelse: Karen-Birgit Bruun, Lone Vigstrup; Sygehus Fyn; Mogens Kjaer-Andersen, Lone Godfredsen, Inge Fauerskov, Pia Bidstrup; Vejle: Flemming Hald-Steffensen, Lise Zeuthen, Elin Petersen, Tove Gohr, Inger Munch, Vibeke Spencer.


The technical assistance from the staff at Department of Clinical Chemistry, Frederiksberg University Hospital is deeply acknowledged. M.S. had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.


View Abstract