Dilated cardiomyopathy

Dilated cardiomyopathy (by which we mean cardiac dilation which is not due to ischaemia or infarction) constitutes the most common primary heart muscle disorder whose natural history may be marked by the development of progressive symptomatic heart failure and major adverse cardiac events including sudden death.6 Traditional markers of risk do not at present adequately predict outcomes,7 and therefore both diagnostic evaluation and risk stratification remain a major challenge. Our group was the first to identify mid-wall LGE (figure 1B) in a significant minority (28%) of patients with dilated cardiomyopathy and to suggest that the presence and amount of this fibrosis is a potential risk factor for adverse events.5 8 This early work suggested a potential prognostic role for LGE. While there was no significant difference in all-cause mortality between patients with and without LGE, our study was underpowered to address this endpoint definitively,8 and significant differences did emerge with respect to a primary composite endpoint of all-cause mortality or cardiovascular hospitalisation (figure 2). These initial positive univariate findings were corroborated by Wu et al who studied a smaller number of higher risk patients with dilated cardiomyopathy.9 In their consecutive prospectively recruited cohort of 65 patients, all of whom had been referred by their treating cardiologist for implantable cardioverter-defibrillator (ICD) implantation and had angiography-verified non-ischaemic cardiomyopathy with left ventricular ejection fraction (LVEF) <35%, a significantly higher proportion of patients with LGE reached the primary composite endpoint of cardiac death, ICD discharge or hospitalisation for heart failure (HR 7.1, 95% CI 2.0 to 25.3; p=0.002).9 Despite studying a higher risk cohort where the event rates may have been expected to be high, given the presence of LGE in only 27 (44%) of the 65 patients studied, in common with earlier work,8 the event rates were too low to permit definitive examination of the independent importance of LGE relative to traditional prognostic markers.

• Download figure
• Open in new tab
• Download powerpoint

Figure 2

Kaplan–Meier survival estimates for a primary composite endpoint of all-cause mortality or cardiovascular hospitalisation in 101 patients with dilated cardiomyopathy stratified according to the presence or absence of late gadolinium enhancement (LGE). Reproduced with permission from Assomull et al.8

More recently, Lehrke et al attempted to evaluate the potential prognostic role of LGE by studying a larger cohort of 184 consecutive patients with non-ischaemic dilated cardiomyopathy.10 They chose the clinically relevant composite endpoint of cardiac death, aborted sudden cardiac death (SCD) (appropriate ICD discharge) and hospitalisation for heart failure. In keeping with previous work, the presence of LGE conferred a significantly increased risk of attaining the primary composite endpoint (HR 3.5, 95% CI 1.36 to 9.02; p=0.01). Intriguingly, subgroup cumulative survival analysis, which was not prespecified, suggested that LGE retained importance only in patients with LVEF <30%. Caution is needed, however, when interpreting these early findings as evidence of independent prognostic significance. The total number of patients experiencing the composite primary endpoint in the LGE group (regardless of underling ejection fraction) was only 15. Stratifying the LGE-positive group further into two subgroups based on LVEF runs the risk of spuriously significant findings due to the related aggregate effects of low event rates and chance. Paradoxically, those without LGE and LVEF <30% in this subgroup analysis appear to have shown a trend towards better survival than those with LVEF >30% and no LGE. A classic and often cited example to illustrate the potential risks and fallacies of exploratory subgroup analysis is the Second International Study of Infarct Survival (ISIS-2) which apparently revealed a significant impact of astrological star sign on the survival benefit of aspirin for myocardial infarction.11

The key central question that remains to be settled is whether LGE is of independent prognostic importance over and above traditional risk factors. Attempts to address this have relied upon multivariate survival analysis using Cox's proportional hazards regression model. When interpreting the results of such analysis, an awareness of the limitations of multivariate techniques in general and the underlying assumptions made by the model is important. All multivariate models seek to explain the variation in a dependent (outcome) variable as a function of a series of independent (predictor) variables or covariates. The accuracy of the estimates of the regression coefficients is driven by the size of the dataset used to construct the model or, in the case of survival analysis, the number of events being examined. If the number of events per covariate is small, there is a high risk of ‘overfitting’ the model to the dataset.12 This leads to unreliable estimates of the magnitude and statistical significance of each covariate. Peduzzi et al13 found that, when the number of events per predictor variable falls to <10, there are highly significant errors in the estimation of the regression coefficients which are used to calculate HRs.13 As an extreme illustration, at two events per predictor variable, they found that the error or bias in the estimates of regression coefficients often exceeded 100%.13

In the study of Lehrke et al, the LGE-positive and LGE-negative groups were significantly different with respect to some important baseline characteristics.10 The LGE-positive group had a significantly lower LVEF, higher LV end-diastolic volume index and LV mass index relative to the LGE-negative group. Patients in the LGE-positive group also had significantly higher levels of brain natriuretic peptide (NT-pro BNP) and there were a significantly higher proportion of patients with New York Heart Association (NYHA) class III symptoms. The presence of LGE could therefore be simply a marker of adverse remodelling and underlying disease severity. However, with only a total of 15 events in their large cohort, in common with previous work, there were insufficient data to examine accurately the independent prognostic importance of LGE controlling for all confounding variables. With only 15 events, optimistically, it is only possible to examine LGE and one other variable, although this would still result in 7.5 events per variable, less than the ideal minimum of 10 to avoid the problems with model validity discussed.13 The question of whether LGE is an epiphenomenon of advanced cardiac remodelling highly correlated with existing prognostic markers, or is in and of itself of independent prognostic importance, therefore remains unanswered.

Hypertrophic cardiomyopathy

Familial hypertrophic cardiomyopathy (HCM) is a principal cause of SCD in young people,14 with an estimated prevalence of 1 in 500 of the general population.15 The clinical challenges that arise in risk stratification of patients with HCM closely mirror those posed by dilated cardiomyopathy. HCM is notable for the marked heterogeneity in both its clinical presentation and natural history.16 This is underscored by the fact that patients who have had aborted SCD or who have died suddenly can be minimally symptomatic,17 18 and that sudden death may be the first presentation of HCM.14 Nevertheless, the majority of the patients with HCM remain asymptomatic and stable without experiencing significant morbidity or mortality.19 However, a small but significant minority experience a more severe clinical trajectory with a significantly increased risk of premature SCD, embolic stroke secondary to atrial fibrillation and death from congestive heart failure.16 20 A number of risk factors for SCD have been identified by observational studies, but these individually have a low positive predictive value and many—such as non-sustained ventricular tachycardia and abnormal blood pressure response to exercise—only appear to be valuable in younger patients.21 These models are likely to be less informative in the increasingly older cohort of patients being identified in contemporary practice, who in turn are more susceptible to the development of heart failure which can affect up to 10% of such individuals.20 22

There is emerging evidence that myocardial fibrosis is a relatively early manifestation of HCM23 and that it may have a pathophysiological role in SCD.24 In particular, it has been suggested that the presence of fibrosis may itself herald a high risk for sudden death in patients without conventional risk factors.25 26 Given its ability to detect fibrosis non-invasively (figure 1C,D), there is therefore a strong pathophysiological rationale for exploring LGE-CMR to facilitate risk stratification in HCM. Early work using this technique has suggested that fibrosis is associated with a propensity towards ventricular arrhythmia27 28 as well as conventional risk markers for SCD.22 Adabag et al found that LGE was an independent risk factor for and conferred a 7.3-fold (95% CI 2.6 to 20.4; p<0.0001) increased RR of non-sustained ventricular tachycardia,27 a potential risk factor for SCD in HCM.29–31 Attempts to define the prognostic importance of LGE for SCD have, however, been constrained by the low event rates seen particularly in community-based HCM cohorts.32

O'Hanlon et al examined the significance of fibrosis as detected by LGE-CMR for the prediction of major clinical events.20 In line with previous findings, approximately two-thirds of patients exhibited detectable replacement fibrosis.22 33 Over a mean follow-up period of 3.1 years, 18.4% of patients reached the prespecified primary composite endpoint of cardiovascular death, unplanned cardiovascular hospitalisation, sustained ventricular tachycardia, ventricular fibrillation or appropriate ICD discharge. The majority of these were in the fibrosis group (figure 3), and Cox proportional hazards modelling revealed the amount of fibrosis to be a significant independent predictor of the primary composite endpoint, conferring a 2.7-fold (95% CI 1.01 to 7.1; p=0.046) increased hazard of adverse events or a HR of 1.15 (95% CI 1.01 to 1.30; p=0.03) per 5% increase in fibrosis (figure 4). However, the difference in the primary outcome between the LGE-positive and LGE-negative groups was driven primarily by hospitalisation for heart failure. There were no significant differences between the groups with respect to cardiovascular mortality or sudden death. A prespecified composite arrhythmic secondary endpoint of sustained ventricular tachycardia, ventricular fibrillation, appropriate ICD discharge or sudden death was also used. In this cohort, 12 out of 217 patients reached this outcome, of whom 10 were in the fibrosis group and two were in a fibrosis-free group. However, with only 12 patients experiencing events over the mean 3-year follow-up period (5.5% of the cohort), the study was underpowered to assess the significance of fibrosis as an independent risk factor for SCD. This event rate is comparable to that found by Maron et al where a comparable composite endpoint also incorporating progressive heart failure was experienced by 5.5% of patients per annum in the LGE group versus 3.3% in the non-LGE group over a mean follow-up period of 1.9 years, a difference which failed to reach statistical significance.33

• Download figure
• Open in new tab
• Download powerpoint

Figure 3

Kaplan–Meier survival estimates in 217 patients with hypertrophic cardiomyopathy for a primary composite endpoint of cardiovascular death, unplanned cardiovascular hospitalisation, sustained ventricular tachycardia/fibrillation or appropriate implantable cardioverter-defibrillator discharge stratified according to the presence or absence of late gadolinium enhancement (LGE). Reproduced with permission from O'Hanlon et al.20

• Download figure
• Open in new tab
• Download powerpoint

Figure 4

Predicted probability at 1, 2 and 3 years of reaching the primary composite endpoint of cardiovascular death, unplanned cardiovascular hospitalisation, sustained ventricular tachycardia/fibrillation or appropriate implantable cardioverter-defibrillator discharge according to the amount of late gadolinium enhancement (LGE). Reproduced with permission from O'Hanlon et al.20

Only one other previous study has attempted to prospectively assess the significance of LGE as an independent predictor of all-cause and cardiac mortality.34 Bruder et al followed up 220 patients with HCM after LGE-CMR imaging over a mean duration of 3 years.34 In their tertiary centre cohort, a total of 16 patients experienced cardiac death, defined as death from all cardiac causes including SCD, heart failure and aborted SCD (ie, appropriate ICD discharge, successful cardioversion or cardiopulmonary resuscitation in a patient who remained alive 28 days thereafter). LGE was seen in 67% of their cohort. The presence of fibrosis appeared to confer an OR of 8.01 (95% CI 1.04 to 61.9) for cardiac death. Interestingly, of the 11 patients who experienced SCD, only three had any conventionally recognised risk factors. Furthermore, when their data were subjected to multivariate analysis, the presence of LGE emerged as an independent predictor of cardiac death whereas the presence of one or two conventional risk factors failed to reach statistical significance. However, once again, owing to the small number of events, they acknowledged that they were unable to subject the cohort with SCD to multivariate analysis. Despite the comparably small number of cardiovascular deaths, they did examine the effect of three covariates (LV ejection fraction, LV mass and the presence of LGE) as predictors of mortality and appeared to find that only the presence of LGE was a significant independent predictor. Similarly, when they examined the impact of the presence of LGE together with one or two risk factors for SCD, the presence of LGE appeared to emerge as the only significant predictor. However, these conclusions are not robust as Cox regression analysis with at least three predictors was used to examine the impact on just 16 events or five events per covariate. They also used univariate analysis to estimate ORs for the impact of LGE on all-cause mortality, calculating a value of 5.47 for all-cause mortality compared with a lower OR of 3.86 for the presence of two traditional risk factors. On the basis of this higher OR and the results of their multivariate analysis, they asserted that LGE may be a better predictor of outcome than the traditional approach of assessing risk factor burden. However, the CIs for the ORs for both LGE and two risk factors were wide and comfortably overlapping at 1.24 to 24.08 and 0.7 to 21.2, respectively. This is despite the relatively high mortality seen in their cohort, which at 9% is higher than that typically seen even in tertiary centre cohorts.16 The use of ORs to compare outcomes in two groups composed of individuals with widely differing lengths of follow-up and, by inference, exposure to risk is also questionable.

In keeping with previous studies, both O'Hanlon et al and Bruder et al found that maximum wall thickness was significantly greater in patients with LGE than in those without LGE.20 34–36 Similarly, as a consequence, all patients with a wall thickness of >30 mm, an established risk factor for sudden death,14 were found in the LGE groups in both studies.20 34 In addition, both O'Hanlon et al and Bruder et al found that LV mass index was significantly higher in the LGE group.20 34 This implies that LGE becomes more prevalent with advanced or more severe disease and may therefore be confounded by other markers of disease severity. Furthermore, as suggested by the work of O'Hanlon et al (figure 4), the amount rather than the mere presence of fibrosis may more accurately reflect the state of the underlying substrate and, in the future, serve as a better marker of risk for cardiovascular mortality. This is reflected by the typical prevalence of LGE in HCM which approaches approximately two-thirds—over an order of magnitude higher than the rate of adverse events experienced by this cohort.

All the studies conducted to date have been underpowered to address whether LGE is an independent predictor of SCD or cardiovascular mortality.20 27 32 34 Significant heterogeneity between studies regarding study design, definition of endpoints, together with statistical analysis and reporting of results, precludes the use of meta-analysis to gain valuable insights into these questions. Future work using statistically robust methodology and large sample sizes needs to focus on addressing variables potentially confounding LGE and to address the importance of the quantity of fibrosis on important hard endpoints.