Increased interleukin-1β levels are associated with left ventricular hypertrophy and remodelling following acute ST segment elevation myocardial infarction treated by primary percutaneous coronary intervention
From the Division of Cardiology, Stavanger University Hospital, Stavanger
Abstract. Ørn S, Ueland T, Manhenke C, Sandanger Ø, Godang K, Yndestad A, Mollnes TE, Dickstein K, Aukrust P (Stavanger University Hospital, Stavanger; Oslo University Hospital Rikshospitalet; University of Bergen, Bergen; University of Oslo; Oslo; Norway). Increased interleukin-1β levels are associated with left ventricular hypertrophy and remodelling following acute ST segment elevation myocardial infarction treated by primary percutaneous coronary intervention. J Intern Med 2012; 272: 267–276.
Objectives. To assess the relationship between interleukin (IL)-1-related molecules, infarct size and left ventricular (LV) remodelling following acute myocardial infarction (MI).
Methods. Forty-two patients with first-time diagnosis of ST segment elevation MI (STEMI), with a single occluded vessel successfully revascularized by primary percutaneous coronary intervention (PCI), were recruited to this observational study conducted at a university teaching hospital and followed for 1 year.
Main outcome measures. Plasma levels of IL-1β, IL-1 receptor antagonist (IL-1Ra), IL-18 and caspase-1 were analysed before and 2 days, 1 week and 2 months after PCI. Serial cardiac magnetic resonance imaging (CMR) was used for the assessment of infarct size and LV remodelling. CMR findings at 1 year was the primary outcome variable.
Results. Univariate analysis showed that IL-1-related mediators were strongly (IL-1 β), moderately (caspase-1) and weakly (IL-1Ra) associated with impaired myocardial function and noninfarct mass, but not infarct size, 1 year after reperfused STEMI. In multivariate analyses, troponin T predicted LV ejection fraction (LVEF), infarct size and LV end-diastolic (LVEDVi) and end-systolic volume index (LVESVi). However, significant additional variance was explained by IL-1β, IL-18 and caspase-1. IL-1β levels at 2 months, IL-18 at 2 days and pre-PCI caspase-1 were predictors of LVEF. Caspase-1 and in particular IL-1β at 2 days were the only predictors of noninfarct mass. IL-1β and IL-18 at 2 days were predictors of LVEDVi, whilst pre-PCI levels of IL-1β contributed to prediction of LVESVi. By contrast, pro-B-type natriuretic peptide, C-reactive protein, IL-6 and transforming growth factor-β1 (TGF-β1) had no or only a weak (TGF-β1) association with these CMR parameters in multivariate analyses.
Conclusions. IL-1β levels after STEMI were strongly associated with impaired myocardial function and noninfarct LV mass after 1 year, suggesting a potential role for IL-1β as a predictor of maladaptive myocardial remodelling following reperfused MI.
Following myocardial infarction (MI), inflammatory mediators are released by the myocardium as a response to tissue injury and contribute to tissue repair and adaptive responses [1–4]. These mediators, however, may also promote tissue damage and myocardial failure through mechanisms such as cardiomyocyte apoptosis and enhanced matrix degradation. Whilst a balanced and transient immune response is most probably adaptive, an imbalanced and nonresolving inflammatory response could be unfavourable leading to maladaptive left ventricular (LV) remodelling and myocardial failure [2–5].
A large number of inflammatory mediators are released during MI, and we and others have shown a relation between the levels of C-reactive protein (CRP), as well as upstream recognition systems (i.e. complement) and upstream mediators (e.g. interleukin [IL]-6), and the degree of myocardial damage following MI [6–9]. Similarly, early changes in circulating levels of chemokines have been shown to correlate with indices of myocardial injury and myocardial function . However, the most important activators and mediators in this complex inflammatory response during MI have not been identified.
The IL-1 family comprises more than 10 cytokines, including IL-1β, IL-1 receptor antagonist (IL-1Ra) and IL-18. Several of these play a critical role in the early inflammatory reaction to invading microbes . However, it seems that the role of IL-1 in inflammation is not restricted to infectious disorders. In response to stimulatory factors that are released during noninfectious tissue damage, the precursors of IL-1β and IL-18 are cleaved into active mature peptides in inflammasomes by the proteolytic enzyme caspase-1, promoting so-called sterile inflammation .
The IL-1 system and its relation to the inflammasome is an important area of research in relation to disorders characterized by sterile inflammation such as ischaemia and reperfusion during MI. Despite its importance, few studies have examined the levels of these parameters during MI in humans. We therefore measured plasma levels of IL-1β, IL-18, caspase-1 and IL-1Ra in 42 patients for 2 months following a first-time diagnosis of ST elevation MI (STEMI). To assess the degree of myocardial damage and dysfunction, serial assessment for 1 year with contrast-enhanced cardiac magnetic resonance imaging (CMR) was performed, allowing reliable determination of the amount of necrotic and non-necrotic tissue, and sequential monitoring during follow-up to assess the healing process . The correlation between IL-1-related molecules and CMR assessment at 1 year following reperfused STEMI was the primary outcome variable.
The present study was designed to study the isolated effects of acute ischaemic injury and reperfusion on infarct healing and LV remodelling and reduce the influence by confounders such as residual ischaemia, activation of proteinases during thrombolytic therapy and previous MI . As systemic markers of myocardial damage are highly dependent on washout in the infarct zone, only patients with first-time STEMI who were successfully treated with primary percutaneous coronary intervention (PCI) were included in the study. Moreover, to allow baseline blood sampling with minimal influence from the ischaemic zone, only patients with an occluded infarct-related artery were included.
Consecutive patients admitted with STEMI and selected for primary PCI were enrolled prospectively at a single centre . The diagnosis of STEMI was defined by typical chest pain and ST segment elevation on electrocardiogram at admission. Patients were included if they (i) had no history of previous MI, (ii) demonstrated acute proximal/middle occluded single-vessel disease, (iii) had undergone successful PCI with stent implantation without significant residual stenosis, (iv) had no contraindications to CMR, and (v) could be scanned within 48 h of PCI. Patients were assessed by measurement of troponin T (TnT) every 24 h during the first 4 days and at the time of the second CMR scan at 7 days to ensure a steady wash-out of the infarcted territory. All patients were treated with aspirin, clopidogrel and a statin. All other medications were prescribed by the treating physician without knowledge of the CMR results. The study was approved by the regional ethics committee of the University of Bergen, Norway, and all patients provided written informed consent prior to inclusion.
All patients underwent scanning four times: at 2 days, 1 week, 2 months and 1 year following PCI. Scans were obtained during repeated breath-holds with patients in a supine position using a 1.5-T whole-body scanner (Intera R 10.3; Philips Medical Systems, Best, the Netherlands) with a dedicated cardiac coil as previously described . Resting LV function was determined with cine images using a steady-state free precession technique (balanced fast-field echo sequence) in short and long axis views in the true heart axis. The short axis covered the whole LV with 10–14 contiguous slices. LV mass and ejection fraction (LVEF) were determined using short-axis volumetry. All postprocessing was performed using View Forum™ Software (Philips Medical Systems). With the exception of LVEF, volumes and mass were indexed to the body surface area: LV end-diastolic volume index (LVEDVi), LV end-systolic volume index (LVESVi), LV stroke volume index (LVSVi) and LV mass index. Infarct mass was assessed manually with planimetry on each short-axis slice, delineating the hyperenhanced area and adding all slices to generate infarct volume. The same density (1.05 g cm−3) was assumed for both hyperenhanced (infarcted) and nonhyperenhanced (noninfarcted) myocardium. Noninfarcted myocardial mass index was calculated by subtracting the infarcted myocardial mass index from the LV mass index.
Blood sampling protocol
Venous blood samples were collected on admission to the hospital, immediately prior to PCI (pre-PCI), and 2 days, 1 week and 2 months following hospitalization. The pyrogen-free blood collection tubes were immediately immersed in melting ice (in ethylenediaminetetraacetic acid [EDTA]-containing tubes to prepare plasma) or maintained at room temperature (in tubes without any additives to prepare serum) and centrifuged within 20 min at 2500 g for 20 min to obtain platelet-poor plasma or centrifuged at 1000 g for 10 min after coagulation to obtain serum. All samples were stored at −80 °C and thawed less than three times.
Plasma levels of IL-1Ra, IL-18 and caspase-1 were determined using enzyme immunoassay (EIA) kits obtained from R&D Systems (Minneapolis, MN, USA). Plasma levels of IL-1β and IL-10 were determined using a multiplex cytokine immunoassay (Bio-Plex; Bio-Rad Laboratories, Hercules, CA, USA) analysed on a Multiplex Analyzer (Bio-Rad Laboratories). The upper limit of IL-1β in 10 healthy controls was 2.46 pg mL−1 as assessed by measurements in our laboratory. The concentration of transforming growth factor β1 (TGF-β1) was measured by EIA (R&D Systems) as previously described . Concentrations of IL-6, CRP, TnT and N-terminal pro-B-type natriuretic peptide (NT-proBNP) have previously been measured in this study population . The intra- and interassay coefficients of variation were <10% for all assays.
Continuous variables are expressed as mean ± standard error of mean (SEM). Time-dependent changes were assessed using the Friedman test. Biochemical parameters were assessed for characteristics of distribution using the Kolmogorov–Smirnov test. Because we were conducting linear regression analysis, all non-normally distributed variables were log transformed prior to statistical analysis, but may be presented nontransformed. Pearson’s correlation was used to investigate associations between variables. A stepwise linear regression was performed to assess the independent contribution of all plasma parameters on CMR indices. A two-tailed P <0.05 was considered significant. However, as multiple comparisons were performed, particular attention should be paid to P-values < 0.01 and correlations at several time-points.
CMR parameters during follow-up
The baseline characteristics of the 42 patients who underwent successful revascularization are shown in Table 1. Briefly, complete reperfusion (TIMI 3 flow) was achieved in 35 patients, whereas seven patients had a slight impairment in coronary flow (TIMI 2 flow) following revascularization, The first CMR scan (2 days) was performed at a mean of 2.2 ± 0.4 days, the second (1 week) at 7.3 ± 0.8 days, the third (2 months) at 61.0 ± 3.7 days and the fourth (1 year) at 364.0 ± 1.2 days following primary PCI. The changes in CMR parameters during 2 months of follow-up have previously been reported . In the present study, these data have been extended to 1 year (Fig. 1). As shown, there was a gradual decrease in infarct size and LVESVi from 2 days to 1 year after PCI, with no changes in LVEDVi during follow-up. There was a gradual increase in LVEF until 2 months after PCI, with no further increase at 1 year (Fig. 1). Noninfarcted myocardial mass showed a transient increase with maximum levels at 1 week, followed by a subsequent and gradual decrease (Fig. 1). In the present study, reperfusion was achieved at about 4.5 h after the onset of symptoms. In accordance with previous animal models , the late timing resulted in significant reperfusion damage and it is therefore not surprising that some patients developed large infarcts and substantial LV remodelling 1 year after PCI (Fig. 1).
Table 1. Baseline characteristics of the study group
ACEI, angiotensin-converting enzyme inhibitor; ARB, angiotensin receptor blocker; Gp, glycoprotein; PCI, percutaneous coronary intervention. Where indicated, data are given as mean ± SD.
58 ± 12
Body surface area (m2)
2.0 ± 0.2
Systolic blood pressure (mmHg)
140 ± 27
Diastolic blood pressure (mmHg)
84 ± 20
Heart rate (beats min−1)
74 ± 17
Time from symptoms to reperfusion (min)
261 ± 202
Q-waves prior to PCI
Q-waves at discharge
Left anterior descending artery
Right coronary artery
Lesion location (proximal/mid/distal)
Patients with more than one stent
TIMI flow post-PCI (3/2)
Gp IIb/IIIa antagonists
Plasma levels of cytokines following PCI
When analysing plasma levels of IL-1-related molecules during 2 months of follow-up, several patterns were revealed (Fig. 2). First, the initial IL-1β levels (pre-PCI) were elevated as compared to healthy controls (upper limit in 10 healthy controls was 2.46 pg mL−1), indicating early activation of IL-1β during STEMI. Secondly, there was a rapid decline in IL-1β from baseline to 2 days after PCI, followed by a subsequent increase reaching levels comparable to baseline after 2 months. However, compared with healthy controls, persistently raised levels were seen throughout the observation period. Thirdly, there was a gradual and significant decline in IL-1Ra throughout the follow-up period. Fourthly, IL-18 showed only minor changes during follow-up, whereas an initial rapid and significant decline in caspase-1 was observed (Fig. 2).
The relationship between IL-1-related molecules and CMR parameters at 1 year: univariate analyses
Cardiac magnetic resonance imaging parameters after 1 year of follow-up are regarded as the most important means of assessment of long-term myocardial damage following MI. We therefore examined the relationship between plasma levels of IL-1-related mediators during the first 2 months and CMR data after 1 year (Table 2). First, IL-1β was positively correlated with LVESVi and LVEDVi at all time-points. Secondly, IL-1β was also negatively correlated with LVEF (at 2 days and 2 months) and positively correlated with noninfarct LV mass (at 2 days and 1 week), but not with infarct size. Thirdly, caspase-1 was negatively correlated with noninfarct LV mass (at 2 days and 2 months) and positively correlated with LVEF (pre-PCI). Fourthly, except for a negative correlation between IL-1Ra (2 days) and LVEF, IL-1Ra and IL-18 were not correlated with any of the CMR parameters after 1 year of follow-up. We also assessed the relationship between changes in IL-1-related parameters and CMR parameters after 1 year, but we did not find that this type approach provided any additional information (data not shown).
Table 2. Correlations between early measurements (pre-PCI to 2 months) of IL-1-related mediators and CMR findings at 1 year following primary PCI
CMR, cardiac magnetic resonance imaging; LVEDVi, left ventricular end-diastolic volume index; LVEF, left ventricular ejection fraction; LVESVi, left ventricular end-systolic volume index, PCI, percutaneous coronary intervention. All plasma variables and infarct mass were non-normally distributed and log transformed prior to analysis. *P <0.05, **P <0.01.
The relationship between IL-1-related molecules and CMR parameters at 1 year: multivariate analyses
Our findings from univariate analyses suggested that IL-1-related mediators and in particular IL-1β were associated with impaired myocardial function and noninfarct mass, but not with infarct size, 1 year after reperfused STEMI. TnT and NT-proBNP and to some extent also CRP have been regarded as important parameters of myocardial damage following MI. We therefore examined the association between IL-1-related mediators, TnT, NT-proBNP and CRP, and CMR parameters after 1 year in a stepwise forward regression model (Table 3). As expected, LVEF, infarct size, LVEDVi and LVESVi were largely predicted by TnT levels at 2 days. However, significant additional variance was explained by IL-1-related molecules for different parameters. Specifically, IL-1β levels at 2 months, IL-18 at 2 days and pre-PCI caspase-1 were predictors of LVEF at 1 year. Notably, caspase-1 and in particular IL-1β levels at 2 days were the only predictors of noninfarct LV mass at 1 year. Moreover, IL-1β and IL-18 at 2 days were predictors of LVEDVi, whilst pre-PCI levels of IL-1β contributed to the prediction of LVESVi. Finally, whereas IL-1β was negatively associated with LVEF and positively associated with noninfarct LV mass, the opposite pattern was seen for caspase-1, consistent with the results of the univariate analyses (Table 2). In contrast to IL-1-related parameters and TnT, NT-proBNP and CRP had no significant effect in this regression model (data not shown).
Table 3. Forward stepwise regression showing predictors (given as nonstandarized coefficient) of CMR indices at 1 year
CMR, cardiac magnetic resonance imaging; LVEDVi, left ventricular end-diastolic volume index; LVEF, left ventricular ejection fraction; LVESVi, left ventricular end-systolic volume index;PCI, percutaneous coronary intervention; TnT, troponin T. All IL-1-related mediators and infarct mass were non-normally distributed and log transformed prior to analysis. *P <0.05, **P <0.01, ***P <0.001.
TnT (2 days)
IL-18 (2 days)
IL-1β (2 months)
Infarct mass (constant)
TnT (2 days)
Noninf. mass (constant)
IL-1β (2 days)
Caspase-1 (2 days)
TnT (2 days)
IL-1β (2 days)
IL-18 (2 days)
TnT (2 days)
The relation between IL-1-related molecules, other inflammatory and anti-inflammatory cytokines and CMR parameters after 1 year
The cytokines operate in a network and the levels of other inflammatory cytokines could therefore influence the interaction between IL-1-related molecules and post-STEMI remodelling . In particular, the interaction between IL-1β and growth factors related to matrix remodelling such as TGF-β1 could be of particular importance .
There was an increase in plasma levels of TGF-β1 following PCI, which reached statistical significance after 7 days (Fig. 3). However, no significant associations were observed between TGF-β1 levels during the first 2 months and CMR data at 1 year, except for an association between TGF-β1 at 2 months and LVEDVi (r = 0.33, P =0.037). When TGF-β1 at 2 months was included in the regression model for LVEDVi as presented in Table 3, it replaced IL-18 at step 3 at 2 days giving an increase in R2 of 0.07 (P =0.005). However, TGF-β1 had no impact on any of the other regression models for CMR parameters as outlined in Table 3. Except for a significant association between caspase-1 and TGF-β1 at 2 days (r = 0.32, P =0.037), we found no association between IL-1-related molecules and TGF-β1 at the same time-points. However, baseline (pre-PCI) levels of IL-1β were associated with TGF-β1 at 1 week (r = 0.35, P =0.04) and 2 months (r = 0.48, P =0.004), suggesting some interaction between IL-1β and TGF-β1. When comparing IL-6 levels with IL-1-related molecules at the same time-points, we found correlations between IL-1Ra and IL-6 at 2 days (r = 0.53, P <0.001) and 1 week (r = 0.40, P =0.01), but not with any of the other IL-1-related molecules. Moreover, when the IL-6 data were included in the present CMR analyses, we found no significant associations between IL-6 levels during the first 2 months and CMR data at 1 year. Furthermore, IL-6 did not affect any of the regression models for CMR parameters as presented in Table 3. With regard to IL-10, the majority (>90%) of the samples were below the estimated detection limit of the assay, and these data were therefore not included in the statistical analyses in relation to CMR measurements.
Inflammatory mediators released during MI may contribute to both adaptive and maladaptive responses; however, the relative importance of different inflammatory factors in these processes is far from clear. In the present study, we have shown that there is an early increase in IL-1β levels prior to PCI, which remains elevated during 1 year of follow-up. IL-1β levels during the first 2 months following reperfused STEMI are strongly associated with impaired myocardial function 1 year after MI. Moreover, caspase-1 and in particular IL-1β at 2 days post-MI are strong predictors of noninfarct LV mass after 1 year. IL-1β is an early cytokine mediator in the innate immune response. Following infection and tissue damage, IL-1β is associated with the release of danger associated molecular patterns . Whereas the release of IL-1β may contribute to the repair process following MI, our findings suggest a link between this inflammatory cytokine and maladaptive myocardial remodelling following reperfused STEMI, which could result in impaired myocardial function and heart failure.
Increased LV mass is a powerful predictor of cardiovascular morbidity and mortality, and the relationship between LV mass and LV volumes provides increased prognostic information beyond either parameter alone [12, 17]. Amongst patients surviving the acute phase following MI complicated by LV dysfunction or heart failure, specific patterns of LV geometry have been associated with death or hospitalization because of heart failure, also after adjusting for LVEF and LVEDV . In the present study, we have demonstrated a positive association between IL-1β levels during STEMI and noninfarcted mass at 1 year. This finding may suggest a potential role for IL-1β not only in the repair process, but also in the regulation of maladaptive trophic responses following MI. Previous studies in experimental models of heart failure have shown that IL-1β may contribute to pressure-mediated hypertrophy, possibly through the involvement of mechanical stretch-induced IL-1β release . The results of in vitro studies and some studies in animal models have suggested that IL-1β may promote maladaptive hypertrophic responses after MI through mechanisms such as stimulation of fibroblast growth and migration, augmentation of angiotensin-II-mediated hypertrophy in cardiomyocytes and promotion of enhanced matrix metalloproteinase activity within the myocardium . Our findings indicate that such mechanisms may also operate in humans. Furthermore, such a notion is supported by a recent small-scale study, suggesting that intervention with Kineret (a synthetic IL-1Ra) is associated with improvement in myocardial function following MI . Taken together, these findings may indicate that the release of IL-1β during MI promotes a maladaptive net effect within the myocardium. However, because our study was not designed to assess clinical outcomes, we cannot be certain that the observed association between IL-1β and LV noninfarcted mass and indices of impaired myocardial function truly represents an adaptive or a maladaptive response.
During MI, several mediators (including cytokines) interact, resulting in a net effect on myocardial remodelling . We have previously reported a marked increase in IL-6 following PCI, which was associated with infarct size after 2 days . However, in contrast to the association between IL-1β and noninfarcted mass, we found no significant associations between IL-6 levels during the first 2 months and CMR data at 1 year. These findings suggest that IL-6- and IL-1-related molecules may, at least partly, reflect different inflammatory pathways during remodelling after reperfused STEMI. The interaction between inflammatory cytokines, including IL-1β, and matrix remodelling is a major feature of myocardial remodelling following MI . In the present study, we have shown an increase in TGF-β1 levels following PCI, but with only minor associations with CMR parameters after 1 year. However, baseline levels of IL-1β were significantly associated with TGF-β1 levels during follow-up, suggesting some interaction between these mediators during myocardial remodelling following reperfused STEMI. The correlation coefficients in the multivariate models were apparently more robust than in the univariate analyses, suggesting that multiple factors are involved in myocardial remodelling following reperfused STEMI. Our findings also suggest that IL-1β and related molecules reflect the contribution of certain pathways that are not accounted for by the traditional markers (i.e.TnT, NT-proBNP and CRP). In this regard, it also underscores the importance of performing multivariate analysis to determine the impact of a novel factor in addition to traditional factors.
Although studies have shown that IL-1β may have a role in atherothrombotic disease and adverse remodelling following MI, few have investigated the relation between IL-1β and myocardial remodelling in relevant patient populations, possibly because of low circulating levels. IL-18, another member of the IL-1 cytokine family, has also been investigated in relation to maladaptive myocardial remodelling  and was recently shown to be an independent predictor of short- but not long-term adverse clinical events in STEMI patients . In the present study, the association with indices of maladaptive remodelling was much stronger for IL-1β than for IL-18 and, importantly, we were able to detect measurable IL-1β levels in all patients with the sensitive multiplex assay. Caspase-1 converts the pro-forms of IL-1β and IL-18 to their active forms , and here we found an association between plasma levels of caspase-1 and parameters of myocardial remodelling. Somewhat surprisingly, however, whereas IL-1β was negatively associated with LVEF and positively associated with noninfarct LV mass, the associations were reversed for caspase-1. The reason for this remains unclear, but it has been suggested that plasma levels of caspase-1 in cardiovascular disorders may, at least in part, reflect pathways other than caspase-1 activity within the myocardium .
The present study has some limitations. The first of these is the relatively small sample size. However, the rigorous selection criteria to reduce the number of confounders, such as prior MI, multivessel disease, differences in revascularization techniques, varying patency in the infarct-related artery prior to and after PCI, and various comorbidities that could involve inflammatory mechanisms, collectively counteract this limitation. Secondly, correlations do not necessarily indicate a causal relationship, in particular when making multiple comparisons as in the present study, and further mechanistic studies are needed to elucidate the role of IL-1-related molecules in post-MI remodelling. Thirdly, systemic cytokine levels may not necessarily reflect their local concentrations within the myocardium, and our data should therefore be interpreted with some caution. Fourthly, the same density (1.05 g cm−3) was assumed for both hyperenhanced (infarcted) and nonhyperenhanced (noninfarcted) myocardium. This is the conventional way to assess the mass of noninfarcted and infarcted myocardium, but the specific gravity differs according to both timing following MI and the type of reperfusion performed . In the present study, however, we found no relationship between the volume of infarcted myocardium and IL-1-related molecules, suggesting that the current data are unlikely to be compromised by potential differences in tissue gravity. Fifth, although the present study was not designed to compare STEMI patients with healthy controls, it should be noted that the upper limit for IL-1β in healthy controls, which was used as a comparator, was based on a small number of controls. Finally, although we compared IL-1-related molecules with plasma levels of IL-6 and TGF-β1, data regarding other mediators in the cytokine network could also be of interest.
In conclusion, our data suggest that IL-1β could be a marker for adverse remodelling and dysfunction after reperfused STEMI, reflecting additional pathways other than, for example, IL-6 and CRP, and potentially providing additional prognostic information. These issues need to be further investigated in larger studies with clinical end-points. One problem, however, is that IL-1β circulates at very low levels. An important challenge is therefore to identify even better markers for the activity in the IL-1 system.
Conflict of interest statement
No conflict of interest to declare.
The authors wish to thank Torbjørn Aarsland and Jorunn Nielsen for their important contributions during the study.
This work was supported by Helse Vest, the Norwegian Association of Heart and Lung Patients, the University of Oslo, the Norwegian Council on Cardiovascular Disease, the Family Blix Foundation and Helse Sør-Øst.