Author + information
- Received October 8, 2015
- Revision received February 22, 2016
- Accepted February 25, 2016
- Published online December 1, 2016.
- Zhen Weng, PhDa,
- Jialu Yao, MDb,
- Raymond H. Chan, MD, MPHc,
- Jun He, MDa,
- Xiangjun Yang, MD, PhDb,
- Yafeng Zhou, MD, PhDb,∗∗ ( and )
- Yang He, MDa,∗ ()
- aCyrus Tang Hematology Center and Ministry of Education Engineering Center of Hematological Disease, the First Affiliated Hospital, and the Collaborative Innovation Center of Hematology, Soochow University, Suzhou, China
- bDepartment of Cardiology, the First Affiliated Hospital of Soochow University, Suzhou, China
- cDivision of Cardiology, Peter Munk Cardiac Center, Toronto General Hospital, University Health Network, University of Toronto, Toronto, Ontario, Canada
Objectives The aims of this study included performing a meta-analysis of the predictive value of late gadolinium enhancement (LGE) cardiac magnetic resonance (CMR) for adverse events and death in hypertrophic cardiomyopathy (HCM).
Background CMR with LGE can identify areas of myocardial fibrosis; however, controversies remain regarding the independent prognostic importance of LGE-CMR in HCM.
Methods We searched PubMed and Web of Science for studies that investigated the prognostic value of LGE in patients with HCM. Pooled odds ratios (ORs), hazard ratios (HRs), and 95% confidence intervals (CIs) were calculated to assess the role of LGE CMR in the risk stratification of HCM.
Results Seven studies were retrieved from 393 citations for the analysis, of which 2 were eliminated because of overlapping data. In total, 2,993 patients (mean age 54.6 years; median follow-up 36.8 months) were included in the analysis. Meta-analysis showed the presence of LGE was associated with an increased risk for sudden cardiac death (SCD) (OR: 3.41; 95% CI:1.97 to 5.94; p < 0.001), all-cause mortality (OR: 1.80, 95% CI: 1.21 to 2.69; p = 0.004), cardiovascular mortality (OR: 2.93, 95% CI: 1.53 to 5.61; p = 0.001), and a trend for heart failure death (OR: 2.21, 95% CI: 0.84 to 5.80; p = 0.107). Extent of LGE was associated with an increased risk of SCD (HR: 1.56/10% LGE; 95% CI: 1.33 to 1.82; p < 0.0001), heart failure death (HR: 1.61/10% LGE; 95% CI: 1.21 to 2.13; p = 0.001), all-cause mortality (HR: 1.29/10% LGE; 95% CI: 1.09 to 1.51; p = 0.002), and cardiovascular mortality (HR: 1.57/10% LGE; 95% CI: 1.30 to 1.89; p < 0.001). After adjusting for baseline characteristics, the extent of LGE remained strongly associated with the risk of SCD (HRadjusted: 1.36/10% LGE; 95% CI: 1.10 to 1.69; p = 0.005).
Conclusions Quantitative LGE by CMR exhibited a substantial prognostic value in SCD events prediction, independent of baseline characteristics. Assessment of LGE can be used as an effective tool for risk stratifying patients with HCM.
- cardiac magnetic resonance
- hypertrophic cardiomyopathy
- late gadolinium enhancement
- myocardial fibrosis
Hypertrophic cardiomyopathy (HCM) is characterized by inappropriate myocardial hypertrophy and a nondilated left ventricle (1,2). Most patients with HCM have an excellent prognosis; however, a significant minority experience adverse events, such as heart failure (HF) and sudden cardiac death (SCD) (3). Furthermore, HCM is the leading cause of SCD in young patients, particularly in young athletes (4,5). Current strategies for identifying high-risk patients are incomplete. Cardiac magnetic resonance (CMR), with the advantages of high spatial resolution and tomographic imaging capability, has been applied in diagnosis and prognosis of HCM (6–8). Besides routine capacity in cardiac structure and function evaluation, CMR with late gadolinium enhancement (LGE) can identify areas of myocardial fibrosis where life-threatening ventricular tachyarrhythmias originate (9). However, controversies remain regarding the independent prognostic importance of LGE-CMR in HCM.
Recently, there has been an increasing interest on the prognostic value of LGE for adverse clinical outcomes. A recent meta-analysis by Green et al. (10) showed significant associations between the presence of LGE and cardiovascular mortality, HF death, and all-cause mortality in HCM. However, at the time of Green’s meta-analysis (10), there were only 4 moderately sized studies investigating LGE and adverse events. In the past 3 years, 3 studies (2 updates and 1 new) with more than 2,400 patients on this topic were published (11–13); thus, we have aimed to perform a meta-analysis to identify the predictive value of LGE for SCD, HF death, all-cause mortality, and cardiac mortality.
Literature search, selection, and data collection process
Papers regarding myocardial fibrosis detected by LGE-CMR in patients with HCM that were published on PubMed and Web of Science until May 2014 were included. The search terms include late gadolinium enhancement, cardiac magnetic resonance, hypertrophic cardiomyopathy, and prognosis. Inclusion criteria were as follows: full-text English publication; studies included patients with HCM who were diagnosed with myocardial fibrosis detected by LGE-CMR; included specified endpoints are described in the following section.
Data abstraction and analysis was performed by 2 researchers independently (Z.W. and J.Y.) and reported on standardized forms. All-cause mortality, cardiac death, SCD/aborted SCD, and HF death were assessed as clinical outcome measures. In cases in which endpoints were not specifically reported, study investigators were directly contacted to obtain raw data. Moreover, we extracted the patient characteristics data as described later in the article. In addition, we reclassified patients with implantable cardioverter-defibrillator (ICD) discharge into aborted sudden death according the data provided by Hen et al. (12).
Study quality was evaluated by modified form of the Newcastle-Ottawa Quality Assessment Scale for Cohort Studies (14), in which the quality of the selected trials was determined on the basis of selection of the study groups and ascertainment of the outcome of interest.
For the meta-analysis, pooled odds ratios (ORs), hazard ratios (HRs), and 95% confidence intervals (CIs) were calculated using a fixed effects model according to the Mantel-Haenszel method. The chosen model was based on the I2 statistic, which showed a lack of significant heterogeneity for the chosen endpoints.
Publication bias was tested using Begg’s funnel plot and Egger’s test (15). A publication bias would be detected if the funnel plot was asymmetric and Egger’s test had a p < 0.05. To determine the relative impact of each study on the pooled results and between-study variability, a “1-study removed” sensitivity analysis was performed. We performed the analyses using Comprehensive Meta Analysis version 2.0 (Englewood, New Jersey) and STATA version 12.0 (Stata Corporation, College Station, Texas). The study is reported according to Preferred Reporting Items for Systematic Reviews and Meta-Analyses guidelines (16).
Search results and study quality
The final search on May 31, 2015, resulted in 393 articles. The majority of the articles were excluded because of unrelated topic, animal studies, or if they were duplicated or a review or commentary article; this resulted in a total of 22 included articles. After full-text screening, 7 articles remained, of which 2 (17,18) were excluded in the final analysis because of overlapping data reported in longer follow-up studies involving the same population. Hence, for the calculation of the pooled ORs, 5 articles (11–13,19,20) were included in this meta-analysis, comprising 2,993 patients, 1,658 (55%) of whom were LGE positive. The review process is depicted in Figure 1 (16).
The average number of patients per study was 599 ± 428, whereas the median was 424 patients (range 202 to 1,293 patients). Median follow-up duration was 36.8 months (range 21.8 to 42.6 months). Most of the studies were single-center prospective or retrospective studies except the study by Chan et al. (11), which was multicenter prospective. The scar assessment methods, including population and quality assessment scores, are listed in Tables 1 and 2.
The age of the patients ranged from 42 to 59 years, and male patients dominated in all of the studies. LGE positivity was found in 1,658 patients who underwent CMR. Diabetes mellitus and atrial fibrillation were the most common comorbidities in these studies. The data on New York Heart Association functional class I to IV, left ventricular ejection fraction (LVEF), wall thickness >30 mm, syncope, sustained ventricular tachycardia/fibrillation, and family history of HCM and SCD are listed in Table 3.
A standard meta-analysis was performed for the outcomes SCD/aborted SCD (5 studies), HF death (4 studies), all-cause mortality (3 studies), and cardiovascular mortality (5 studies) because sufficient data were available. The detailed results of our meta-analysis are shown in Figure 2.
The study by Chan et al. (11) demonstrated the presence of LGE was independently associated with a higher risk for SCD (HRadjusted: 2.56; p = 0.02). They also showed that the amount of LGE was both a strong univariate (unadjusted HR: 1.50/10% LGE; p = 0.0001) and multivariable predictor (HRadjusted: 1.46/10% LGE; p = 0.002) for sudden death events. The study by Hen et al. (12) showed that presence of LGE was also independently associated with cardiovascular events (HRadjusted: 7.43; p = 0.05) in multivariable analysis. In contrast, Ismail et al. (13) showed the presence of LGE to only have a trend toward increased risk of SCD events (unadjusted HR: 2.69; p = 0.07). The amount of LGE was not an independent predictor of SCD risk (HRadjusted: 1.10/5% LGE or 1.21/10% LGE; p = 0.299), but was a univariate predictor of SCD risk (unadjusted HR: 1.24/5% LGE or 1.53/10% LGE; p = 0.007). The extracted outcomes of interest for all these studies are listed in Table 4.
Cumulative meta-analysis, which includes subsequently overlapped results from Maron et al. (17) and O’Hanlon et al. (18), showed an increased risk of SCD/aborted SCD associated with the presence of LGE becoming evident after the study by Chan et al. (11), when the pooled OR became 2.98 (95% CI: 1.66 to 5.34; p < 0.001) at that point. After including all studies to date, the pooled OR was 2.92 (95% CI: 1.76 to 4.84; p < 0.001) (Figure 3A). After excluding the overlapped data, the association between LGE and SCD became even stronger (pooled OR: 3.41; 95% CI: 1.97 to 5.94; p < 0.001) (Figure 3B).
Two studies in the meta-analysis had much larger sample sizes than the other studies; hence, we performed a 1-study-removed analysis to assess whether the removal of any one study changed the overall results (Online Figure 1). No change was detected.
Only 3 publications reported risk ratios expressed by quantitative LGE by % left ventricular (LV) mass (11,13,19). These results, supplemented by unpublished data from Chan et al. (11), are summarized in Table 5. Univariate analysis showed a relationship between the amount of LGE and the risk of SCD (pooled HR: 1.56/10% LGE; 95% CI: 1.33 to 1.82; p < 0.001). Adjusted HRs were only reported in 2 studies. The extent of LGE remained a strong independent predictor for SCD (pooled HRadjusted: 1.36/10% LGE; 95% CI: 1.10 to 1.69; p = 0.005), even after adjustments for baseline characteristics. Thus, when compared with patients without LGE, the HRadjusted of SCD related to %LGE is as follows: 10%, HRadjusted: 1.36; 15%, HRadjusted: 1.59; and 20%, HRadjusted: 1.86 (Table 6).
When the cumulative meta-analysis was repeated for HF death, Figure 4A to 4C shows the undue influence of the study by Ismail et al. (13) in 2014. After Chan et al. (11) was reported, there appeared to be a significant relationship between LGE and risk of HF death (pooled OR: 7.29; 95% CI: 1.68 to 31.59; p = 0.008). Indeed, up to that point, all 18 HF deaths occurred in patients with LGE (Table 4). However, Ismail et al. (13) reported 4 of the 14 patients with HF deaths had no LGE, which led to a nonsignificant relationship between LGE and HF death (pooled OR: 2.21; 95% CI: 0.84 to 5.80; p = 0.107) (Figure 4, Online Figure 1). However, when this risk was expressed as quantitative LGE, there was a highly significant increase in risk for HF mortality with increasing amounts of LGE (pooled HR: 1.61/10% LGE; 95% CI: 1.21 to 2.13; p = 0.001). Adjusted analysis was not possible because of the low number of patients with HF death in each study (n ≤ 14).
All-cause mortality and cardiovascular mortality
Similar analysis were performed (both including and excluding overlapped data where applicable) for all-cause mortality (pooled OR: 1.80; 95% CI: 1.21 to 2.69; p = 0.004) and cardiac death (pooled OR: 2.93; 95% CI: 1.53 to 5.61; p = 0.001) (Figures 5 and 6, Online Figure 1), showing a consistently higher risk for such events in patients with LGE. There did not appear to be undue influence of any 1 study in the 1-study-removed analysis (Figures 5 and 6, Online Figure 1). When examining the risk expressed by quantitative LGE, there was also a significant relationship between the amount of LGE and the risk of all-cause mortality (pooled HR: 1.29/10% LGE; 95% CI: 1.09 to 1.51; p = 0.002), and cardiac death (pooled HR: 1.57/10% LGE; 95% CI: 1.30 to 1.89; p < 0.001). Only Chan et al. (11) reported adjusted HRs for all-cause mortality (HRadjusted: 1.51; 95% CI: 1.13 to 2.01; p = 0.006) and cardiac death (HRadjusted: 1.29; 95% CI: 0.93 to 1.77; p = 0.12).
A Begg’s funnel plot for SCD/aborted SCD showed that the studies were equally distributed around the overall estimate (Figure 7). Moreover, the Egger’s test showed no publication bias (p = 0.132).
Because of the variability of phenotypic and clinical presentation of HCM, current strategies do not completely identify all high-risk patients with HCM. There remains a clinical need to identify novel markers to help in risk stratification. Contrast-enhanced CMR has been identified as 1 such candidate, and recently there has been significant interest to assess the prognostic utility of LGE imaging.
In the present meta-analysis of 5 unique studies with 2,993 patients over a median follow-up of 3.1 years, we demonstrate that extensive LGE was a powerful risk marker for SCD, even after adjusting for baseline characteristics, including LVEF. LGE of 20% of LV mass confers an almost 2-fold increase in SCD risk. Thus, not only is the mere presence of LGE associated with SCD (in addition to all-cause mortality and cardiac death), these risks appear to have a positive and continuous relationship with the extent of LGE. These results support the hypothesis that extensive LGE may potentially be used as a prognostic marker to identify high-risk patients with HCM who may be managed more aggressively, including more intensive medical treatment, surveillance, and device implantation.
Our study is consistent and extends those of previous reports. Green et al. (10) performed a meta-analysis in 2012 by evaluating 1,063 patients in 4 studies over an average follow-up of 3.1 years, and their results demonstrated that the presence of LGE was significantly associated with cardiac death, HF death, and all-cause mortality in HCM, and showed a trend toward an association with SCD/aborted SCD. However, there were no data regarding quantitative LGE.
Since that meta-analysis, 3 large studies (11–13) have been published. The meta-analysis by Briasoulis et al. (21) included 2 of these 3 new studies, including 3,067 patients over an average follow-up of 3.1 years. They also found that LGE was associated with an increased risk of SCD, cardiac mortality, and all-cause mortality. They attempted to calculate the risk for adverse events associated with quantitative LGE, but were unable to find a significant relationship using meta-regression techniques. The meta-analysis by Briasoulis et al. (21) failed to take into account the significant number of overlapping subjects in the included studies. Consequently, effect sizes among those studies were positively correlated, thus adversely affecting the validity and precision of the final effect estimates in the meta-analysis.
In contrast, our meta-analysis excluded data from Maron et al. (17) and O’Hanlon et al. (18) in the final calculations, because the patient populations overlap those in Chan et al. (11) and Ismail et al. (13), respectively. These excluded data, however, did provide some useful insights in the cumulative meta-analysis. They demonstrated there is a consistent accumulation of evidence to support that LGE is associated with increased risk. In addition, we had primary access to unpublished data from the largest study by Chan et al. (11) regarding numerous endpoints (summarized in Table 4), leading to more robust effect estimates.
Although the presence of LGE clearly portends a higher risk of adverse events, it is clinically impractical to use that as a binary tool for making clinical decisions (particularly for primary ICD implantation) because up to 70% of all patients with HCM have some degree of LGE on CMR. The data thus infer that most of such patients can be considered potential candidates for ICDs, which would include a substantial proportion of low-risk patients that would not benefit from such therapy. Implicit in this designation of high-risk status is that there are equal levels of risk across the entire spectrum of LGE amount (i.e., from minimal to substantial). Quantification of LGE can thus improve upon this strategy and is far more useful clinically when trying to risk stratify patients. The extent of LGE appears to be a strong independent predictor of SCD, even after adjusting for baseline characteristics. Each 10% increase in LGE was associated with a 36% relative increase in risk for SCD.
There remains some controversy as to whether LGE provides incremental information over traditional risk factors and whether it should be used routinely as part of clinical risk assessment and direct patient management decisions. The current European Society of Cardiology guidelines for management of hypertrophic cardiomyopathy (22) conspicuously omits extensive LGE by quantitative contrast CMR in the HCM risk-SCD score calculations. Thus far, only Chan et al. (11) have been able to show a robust independent relationship between the amount of LGE and risk for sudden death events (HRadjusted: 1.46/10% LGE; p = 0.002). Ismail et al. (13) found that the amount of LGE was only a univariate predictor of SCD risk (unadjusted HR: 1.24/5% LGE or 1.53/10% LGE; p = 0.007), but not an independent predictor of SCD risk after adjusting for LVEF (adjusted HR: 1.10/5% LGE or 1.21/10% LGE; p = 0.299). However, this discrepant observation may be entirely due to the limited number of endpoints (n = 22) observed in the latter study. Furthermore, it appears from their data that the proportion of patients with LV dysfunction suffering SCD events was substantially higher than those with normal LV function, whereas the vast majority (92%) of those with LVEF <50% had some degree of LGE. This may explain why LGE did not remain an independent predictor after adjusting for LVEF. However, by combining the data in these 2 studies, this current meta-analysis demonstrated that extensive LGE was significantly associated with an elevated SCD risk, even after adjustment for potential confounders.
Without access to the original data to both studies, it is difficult to ascertain whether LGE has truly independent prognostic value after adjusting for traditional risk factors, including LVEF. We believe the extent of LGE is going to be a far more clinically useful metric than using LGE positive/negative to predict adverse events. That is, the positive predictive value of LGE is expected to be very low because the prevalence of LGE in HCM patients is so high. These data will need further validation in separate, larger populations with longer follow-up; research is ongoing (23). More information may also be gleaned from novel CMR techniques such as T1 mapping, which could provide further insights into the role of an abnormal myocardial substrate on adverse events.
Using the current data, we can extrapolate graded levels of risk that may serve as a clinically useful tool to estimate relative risk depending on the amount of LGE (Table 6). For example, 20% LGE confers almost double the risk of SCD compared with patients without LGE (pooled HRadjusted: 1.86). There is a continuous relation between %LGE and SCD risk and thus there is no rigid (and indeed arbitrary) cutpoint for %LGE when making decisions regarding ICD implantation. This information about relative risk can be applied to patients whose risk remains ambiguous after consideration of conventional risk factors, whereby extensive LGE can arbitrate ICD decisions. Furthermore, this graded level of risk allows for a measure of individualization regarding what may compose an unacceptable level of risk to the individual patient.
Although we were unable to find a significant correlation between LGE and HF deaths, the reason is primarily because of the inclusion of data from Ismail’s study (13), where no significant difference could be found in HF death risk between LGE positive and negative patients (10 of 471 vs. 4 of 240, p = 0.90). The relative high weight of their study resulted in a nonsignificant relationship between LGE and HF death. It must be noted that the aggregate number of events was small, thus more data will be needed to draw more definitive conclusions. However, data from quantitative LGE does suggest a possible prognostic role of LGE in HF death; however, this is not adjusted for LVEF.
Although it is generally presumed that LGE equates myocardial fibrosis (and as such these terms are commonly used interchangeably), this may not always be the case. LGE, when isolated to the right ventricular insertion areas, may represent merely expanded extracellular volume rather than replacement fibrosis (24). Recently, Chan et al. (25) demonstrated that LGE isolated to right ventricular insertion areas is not associated with increased risk. However, none of the included studies in this meta-analysis provided prognostic data with regards to LGE isolated to these areas.
First and foremost, this current meta-analysis was limited by inconsistent characteristics of the study populations and variability in definitions of the measured outcomes of interest. Statistically combining significantly heterogeneous data may be problematic. The aggregate sample size used in our meta-analysis was limited, particularly with respect to the total number of events for all-cause mortality and HF death. Because we do not have access to all raw datasets, we are unable to control for traditional risk factors that may confound the relationship between LGE and adverse events. All studies excluded high-risk patients who had an ICD because of the incompatibility of these devices with magnetic resonance imaging, potentially biasing the results. Ideally, LGE by CMR would be performed at the time of diagnosis, but that may not be feasible in all patients.
Our meta-analysis demonstrates the risk of SCD to be significantly associated with not only with the presence of LGE, but more importantly, with the extent of LGE, even after adjusting for baseline characteristics. The presence and extent of LGE was also associated with an increased risk of other adverse events. Extensive LGE may thus be potentially considered a novel risk marker to help identify high-risk patients who are candidates for life-saving therapy with ICD.
COMPETENCY IN MEDICAL KNOWLEDGE: LGE by CMR exhibited a substantial prognostic value in SCD events prediction, even after adjustment for baseline characteristics.
TRANSLATIONAL OUTLOOK: Assessment of myocardial fibrosis by LGE can be used as effective tool to risk stratify patients with HCM.
For a supplemental figure, please see the online version of this article.
This work was supported by the Priority Academic Program Development of Jiangsu Higher Education Institutions and National Natural Science Foundation of China (81170174). All authors have reported that they have no relationships relevant to the contents of this paper to disclose. Drs. Weng, Yao, and Chan contributed equally to this work.
- Abbreviations and Acronyms
- confidence interval
- cardiac magnetic resonance
- hypertrophic cardiomyopathy
- hazard ratio
- heart failure
- late gadolinium enhancement
- left ventricular
- left ventricular ejection fraction
- odds ratio
- sudden cardiac death
- Received October 8, 2015.
- Revision received February 22, 2016.
- Accepted February 25, 2016.
- American College of Cardiology Foundation
- Semsarian C.,
- Ingles J.,
- Maron M.S.,
- Maron B.J.
- Maron B.J.,
- Spirito P.,
- Ackerman M.J.,
- et al.
- Maron M.S.,
- Maron B.J.,
- Harrigan C.,
- et al.
- Pennell D.J.
- Reichek N.,
- Gupta D.
- Maron M.S.
- Chan R.H.,
- Maron B.J.,
- Olivotto I.,
- et al.
- Hen Y.,
- Iguchi N.,
- Utanohara Y.,
- et al.
- Ismail T.F.,
- Jabbour A.,
- Gulati A.,
- et al.
- ↵Wells GA, Shea B, O'Connell D, et al. The Newcastle-Ottawa Scale (NOS) for assessing the quality of nonrandomized studies in meta-analyses. Available at: http://www.ohri.ca/programs/clinical_epidemiology/oxford.asp. Accessed July 1, 2016.
- Egger M.,
- Davey Smith G.,
- Schneider M.,
- Minder C.
- Maron M.S.,
- Appelbaum E.,
- Harrigan C.J.,
- et al.
- O'Hanlon R.,
- Grasso A.,
- Roughton M.,
- et al.
- Bruder O.,
- Wagner A.,
- Jensen C.J.,
- et al.
- Rubinshtein R.,
- Glockner J.F.,
- Ommen S.R.,
- et al.
- Briasoulis A.,
- Mallikethi-Reddy S.,
- Palla M.,
- Alesh I.,
- Afonso L.
- Elliott P.M.,
- Anastasakis A.,
- Borger M.A.,
- et al.