Article Text

Download PDFPDF
Optimising the dichotomy limit for left ventricular ejection fraction in selecting patients for defibrillator therapy after myocardial infarction
  1. Yee Guan Yap1,
  2. Trinh Duong2,
  3. J Martin Bland2,
  4. Marek Malik1,
  5. Christian Torp-Pedersen3,
  6. Lars Køber4,
  7. Mark M Gallagher1,
  8. A John Camm1
  1. 1Department of Cardiological Sciences, St George’s Hospital Medical School, London, UK
  2. 2Department of Public Health Science, St George’s Hospital Medical School, London, UK
  3. 3Department of Cardiology, Gentofte University Hospital, Hellerup, Denmark
  4. 4Department of Cardiology, The National Hospital, Copenhagen, Denmark
  1. Correspondence to:
    Associate Professor Dr Yee Guan Yap
    Department of Medicine, Faculty of Medicine and Health Sciences, University Putra Malaysia, 10B Floor, Grand Seasons Avenue, No 72, Jalan Pahang, Kuala Lumpur 53000, Malaysia; ygyap{at}aol.com

Abstract

Background: The selection of patients for prophylactic implantable cardioverter-defibrilator (ICD) treatment after myocardial infarction (MI) remains controversial.

Aim: To determine the optimum left ventricular ejection fraction (LVEF) dichotomy limit for ICD treatment in patients with a history of MI.

Methods and results: Data from the placebo arms of four randomised trials were pooled to create a cohort of 2828 patients (2206 men, mean (SD) age 65 (11) years) with reduced left ventricular function after MI. The median LVEF was 33% (range 6–40%). LVEF significantly predicted mortality. Each 10% reduction in LVEF <40% conferred a 42% increase in all-cause mortality, a 39% increase in arrhythmic cardiac mortality and a 49% increase in non-arrhythmic cardiac mortality over the 2-year period of follow-up (p<0.001 for all modes of mortality). As the LVEF progressively decreased from ⩽40% to ⩽10%, the data show a U-shaped relationship between the dichotomy limit for LVEF used and the number of patients who must be treated to prevent one arrhythmic death in 2 years. At an LVEF of 16–20%, more patients are likely to die from arrhythmic than non-arrhythmic cardiac deaths, whereas in those with LVEF ⩽10% all deaths were non-arrhythmic. However, the total number of deaths substantially decreased with lower LVEF.

Conclusion: A trade-off exists between the sensitivity and positive predictive accuracy across a range of LVEF, and no single dichotomy limit is completely satisfactory. In patients with LVEF ⩽10% ICD treatment was not beneficial as all patients in this subgroup died from non-arrhythmic causes. The use of a single dichotomy limit for LVEF alone is not sufficient in selecting patients for ICD treatment in the primary prevention of cardiac arrest.

  • AMI, acute myocardial infarction
  • DIAMOND-MI, Danish Investigation of Arrhythmias and Mortality On Dofetilide-Myocardial Infarction
  • EMIAT, European Myocardial Infarct Amiodarone Trial
  • ICD, implantable cardioverter defibrillator
  • LVEF, left ventricular ejection fraction
  • MADIT II, Multicenter Automatic Defibrillator Implantation Trial II
  • MI, myocardial infarction
  • TRACE, TRAndolapril Cardiac Evaluation

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

The Multicenter Automatic Defibrillator Implantation Trial II (MADIT II) showed that the prophylactic implantation of a defibrillator in patients with a prior myocardial infarction (MI) and a left ventricular ejection fraction (LVEF) of ⩽30% reduced mortality from 19.8% to 14.2% at 2 years,1 an effect explicable entirely by a 67% reduction of sudden cardiac death.2 In this patient group, 18 high-risk patients needed to be treated for 2 years to save one life. Despite the approval by the US Food and Drug Administration for the use of implantable cardioverter defibrillator (ICD) in patients meeting the MADIT II entry criteria, the cost effectiveness of the use of ICD treatment in this group is still debated widely.

In patients with a history of MI, there is a progressive increase in 1 year mortality as the LVEF falls <40%.3 Subgroup analysis of previous trials suggests that treatment with amiodarone provides most benefit in patients with moderate (LVEF 31–40%) rather than severe left ventricular impairment (⩽30%) who are more likely to die from pump failure.4 The aim of this study was to assess the relationship between LVEF and the risk of arrhythmic death and thus to determine the optimum dichotomy limit for LVEF in selecting patients who had MI whose risk of arrhythmic death is sufficient to justify prophylactic ICD treatment.

METHODS

We pooled the individual placebo patient data from European Myocardial Infarct Amiodarone Trial (EMIAT), Survival With ORal d-sotalol, TRAndolapril Cardiac Evaluation (TRACE) and Danish Investigation of Arrhythmias and Mortality on Dofetilide-Myocardial Infarction (DIAMOND-MI) studies, which recruited patients with LVEF ⩽40% after MI. Details of the design of these studies have been published elsewhere.4–7

All four trials pooled for this study were prospective multicentre randomised placebo-controlled trials conducted in the thrombolytic era, which recruited patients with recently documented acute myocardial infarction (AMI), LVEF ⩽40% and clear clinical end points determined by an event committee including all-cause mortality, arrhythmic cardiac mortality and non-arrhythmic cardiac mortality. Similar rules were employed by each of the committees. Arrhythmic death was defined according to the Hinkle and Thaler classification as “the abrupt spontaneous cessation of respiration and pulse and loss of consciousness in the absence of other progressive severe medical conditions likely to cause death”. Documented arrhythmic death or cardiac arrest was considered to be present if ventricular tachycardia or fibrillation was recorded within 10 min of clinical death. Other cases were classified as “presumed arrhythmic”.

In the selected studies, patients were followed up for at least 24 months. We analysed survival at 2 years in all trials. All the trials selected were antiarrhythmic drug trials involving class III antiarrhythmic agents except TRACE, which investigated the use of the ACE inhibitor trandolapril. As class III antiarrhythmic drugs and trandolapril have been shown to influence mortality from arrhythmia, we used only data pertaining to patients in the placebo arm of each trial.

Short-term and long-term survival were analysed separately to allow for differences between the studies in the interval between the index MI and trial entry. Short-term survival was analysed using logistic regression at 45 days from the time of the index infarct using only data from the TRACE and DIAMOND-MI studies that recruited patients within 2 weeks of infarct. Long-term survival in patients surviving at least 45 days after the index MI was analysed from the combined data of all four trials. The effect of LVEF on long-term survival was investigated using Cox’s regression, adjusted for treatment and study effects and for demographic factors that had been preassessed and shown to be associated with survival. A two-sided significance test was used, with p<0.05 considered significant. We investigated whether hazards remained proportional over time by comparing the results before and after the first year. Non-linear effects of continuous variables were assessed by adding polynomial terms.

To assess the optimum dichotomy limit for LVEF, we examined the number of patients needed to treat to prevent one arrhythmic or sudden cardiac death. The number needed to treat is defined as the number of patients who would have to receive the treatment for one of them to benefit, calculated as one divided by the absolute risk reduction. As we are only dealing with the placebo patients and had no treatment group to compare, we calculated the number needed to treat on the basis of the finding from MADIT II that ICD implantation in this patient group reduces sudden cardiac death by 67% without altering other forms of mortality.1 Patients who were censored before 45 days were excluded from this analysis. Patients who died from other causes (non-arrhythmic cardiac death and non-cardiac death) were included in the total population.

RESULTS

Data pertaining to 2828 placebo patients were pooled (2206 men, aged: 65 (11) years) from all four trials. Table 1 summarises the baseline characteristics of the patients. The median LVEF was 33% (range 6–40%). There were 652 deaths (all-cause mortality) at 2 years, of which 303 were arrhythmic cardiac deaths and 242 were non-arrhythmic cardiac deaths. After adjusting for factors that were associated with survival (ie, age, sex, previous MI, hypertension history, New York Heart Association class, presence of abnormal Q wave in the surface ECG, heart rate, smoking and systolic blood pressure), LVEF significantly independently predicted long-term (45 day–2 year) all-cause mortality, arrhythmic cardiac mortality and non-arrhythmic cardiac mortality (p<0.001 for all three modes of mortality; table 2).

Table 1

 Baseline characteristics of patients

Table 2

 Effect of incremental 10% increase in left ventricular ejection fraction on survival

Each 10% reduction in LVEF <40% conferred a 42% increase in all-cause mortality, a 39% increase in arrhythmic cardiac mortality and a 49% increase in non-arrhythmic cardiac mortality over the 2-year period. The total number of patients and, consequently, the number of deaths with LVEF <20% was much lower than the number with LVEF 31–40%. Patients with LVEF <20% were slightly more likely to die from non-arrhythmic than from arrhythmic causes, whereas the reverse was true if the LVEF was 31–40% (table 3). LVEF was not predictive of short-term mortality up to 45 days after AMI (all-cause mortality, p = 0.1; arrhythmic mortality, p = 0.1; cardiac mortality, p = 0.96).

Table 3

 Number of patients for different intervals of left ventricular ejection fraction and rates of different forms of mortality

As the LVEF dichotomy limit was reduced from ⩽40% to ⩽15%, the number of patients needed to treat to prevent one arrhythmic death in 2 years declined from 9.3 (for LVEF ⩽40%) to 6.1 (for LVEF ⩽25%) but then increased to 7.9 (for LVEF ⩽15%). This trend became more pronounced, showing a progressively deeper U shape as the period of follow-up was reduced to 3 months (fig 1 and table 4).

Table 4

 Number of patients for three different modes of death at various left ventricular ejection fraction dichotomy limits at a follow-up period of 2 years

Figure 1

 The use of various left ventricular ejection fraction LVEF dichotomy limits and different periods of follow-up. The number of patients needed to treat to prevent one arrhythmic death followed a U-shaped curve that is more accentuated at shorter periods of follow-up.

When we examined the number of patients needed to treat to prevent one arrhythmic death in 2 years at different ranges of LVEF (ie, 5%, 10%, 15%, 20%, 25% and 30%) and at various dichotomy limits (⩽10, ⩽15, ⩽20, ⩽25, ⩽30 ⩽35, 0–10, 0–15, 11–15, 0–20, 11–20 and so on), it was lowest at 4.3 when the LVEF was 16–20% (fig 2). Among patients with LVEF ⩽10%, all deaths occurred exclusively from non-arrhythmic cardiac causes.

Figure 2

 Left ventricular ejection fraction (LVEF) is represented in 5% intervals rather than dichotomy limits. The number of patients needed to treat was lowest at 4.3 when the LVEF was 16–20%.

DISCUSSION

The bulk of available evidence regarding indications for ICD treatment as primary prevention is derived from three trials that used a dichotomy limit for LVEF only as a selection criterion. On the basis of these trials, ICD implantation is commonly undertaken on the sole basis of LVEF <30% in patients after MI. Although there is evidence that low-risk and high-risk subgroups exist within this group, there is no agreed way to stratify for risk. Doing so could allow substantial cost saving at negligible sacrifice in population life expectancy.8 There is also uncertainty about the need for ICD treatment in those with an LVEF of 30–40%. Patients in this group do not generally receive ICD treatment, although, as we have shown, their risk of sudden death is substantial.

This study showed that as the LVEF reduced from 31–40% to <20%, the incidence of all-cause and arrhythmic cardiac mortality increased threefold and that of non-arrhythmic cardiac mortality increased fivefold such that the risk of non-arrhythmic cardiac death superseded that of arrhythmic death. In line with this, the number of patients who must be treated to prevent one arrhythmic death in 2 years decreased as the LVEF dichotomy limit decreased until a “trough” was reached when a dichotomy limit of ⩽25% was used.

The number of patients who must be treated to prevent one arrhythmic death reduced to its lowest value of 4.3 in patients with LVEF in the range 16–20%. As the LVEF decreased further, the number of patients needed to treat began to rise. At LVEF ⩽10%, all deaths were from non-arrhythmic cardiac causes (table 3). This suggests that the use of a single cut-off value of LVEF is inappropriate, and that the specific risk of arrhythmic and non-arrhythmic mortality should be estimated for each individual. Refinement of risk measurement is necessary. This could involve the use of improved strategies for combining the available indices or the development of indices more precise than those considered in this study.

Combining LVEF and other criteria

Of the ICD primary prevention trials in patients with ischaemic heart disease, only the Multicenter Unsustained Tachycardia Trial (LVEF ⩽35%, non-sustained ventricular tachycardia and inducible and non-suppressible ventricular tachycardia at electrophysiological study) and coronary artery bypass graft-patch (LVEF <30%, positive signal-averaged ECG and need for coronary bypass surgery) involved the use of LVEF in combination with other risk factors as entry criteria. In the Multicenter Unsustained Tachycardia Trial, fewer patients needed to be treated with ICDs to save one life than in MADIT II,9 but the registry patients from this study who were non-inducible at electrophysiological testing and had an LVEF <30% had a total mortality and arrhythmic mortality nearly identical to that of patients who were inducible but had an LVEF of 30–40%,8 and data from MADIT II suggest that electrophysiological testing adds little prognostic information.10 Thus, patients who had MI with an LVEF <30% remain at risk of sudden death even if electrophysiological testing proves negative. Other methods of risk stratification such as microvolt-level T-wave alternans seem more promising.11

Time dependence of arrhythmic risk

The number of patients needed to treat to prevent one arrhythmic death reduced significantly as the follow-up period increased from 3 months to 2 years. Thus, 3 months of follow-up is inadequate to provide a true picture of mortality after MI, particularly for arrhythmic mortality. This is consistent with the MADIT II study which showed that although approximately 88% of the patients were recruited >6 months from the last MI, prophylactic ICD treatment confers a survival benefit.1

We did not find any effect of LVEF on short-term mortality up to 45 days after AMI. This observation is intriguing in view of the recently published Defibrillator in Acute Myocardial Infarction Trial that showed that patients who had MI with a reduced LVEF ⩽35% and impaired cardiac autonomic modulation did not benefit from ICD treatment if implanted early after MI (within 6–40 days after MI).12 The mechanism for this is unclear but it seems that remodelling of the left ventricle that occurs in the early weeks after MI makes risk assessment based on a single echocardiographic study less reliable than in the chronic phase.13 It may also be that non-arrhythmic causes of death are so frequent in the early months after acute MI that the benefit of ICD treatment is masked, or that the proximity to revascularisation procedures reduces the benefit of ICD treatment.14

Cost effectiveness

The evidence from this study highlights the heterogeneity of the risk of sudden death in those with LVEF ⩽40%. Patients with an LVEF of 16–20% seem likely to benefit most from ICD treatment, although these account for only a small minority of the sudden deaths in the entire cohort (33 of 294 sudden deaths; table 4). Our study showed that patients with LVEF⩽10% will not benefit from ICD treatment since all patients in this subgroup died from non-arrhythmic causes (fig 2).

The cost effectiveness of ICD use in addition to optimal medical treatment is dependent on absolute risk of arrhythmic death and on the ratio of sudden to non-sudden cardiac death.15 In patients with high rates of sudden cardiac death but low rates of non-sudden death, the ICD provides a large benefit at a reasonable cost. An increase in the rate of deaths not preventable by the ICD within the lifetime of the device would erode greatly its cost-effectiveness.15 Further studies are needed to determine the cost-effectiveness of ICD use among patients who had MI with low LVEF.

LIMITATIONS

The modes of death were determined by event committees, and the exact mode of death cannot be known for certain in all cases. This method of classifying the mode of death has proved to be consistent and sufficiently accurate in the ICD trials. We are dependent on a single measurement of LVEF in most patients, and measurement is operator dependent and often variable over time, particularly in the period soon after MI.

The proportion of patients receiving thrombolytic treatment was low in our study compared with currently expected standards and the data pertain to a period before the widespread use of percutaneous revascularisation, glycoprotein IIb/IIIa receptor inhibitors or clopidogrel in acute MI. The use of ACE inhibitors and β-blockers in the study population was also low compared with the use observed in more recent trials including MADIT II. The widespread use of ACE inhibitors and β-blockers could reduce the rate of both arrhythmic and non-arrhythmic death and alter the balance between arrhythmic and non-arrhythmic deaths in a manner that is difficult to predict. The use of statins could also influence both arrhythmic and non-arrhythmic mortality,16 but their use was not recorded in all the trials analysed. The use of all appropriate pharmacological treatments and cardiac resynchronsation could reduce the rate of non-arrhythmic mortality in those with low LVEF sufficiently to allow them to benefit from the reduction in the risk of arrhythmic mortality provided by ICD treatment. The population of patients in MADIT II was more stable than that in our study in terms of the risk of sudden cardiac death, as almost 90% patients enrolled in MADIT II experienced MI >6 months earlier compared with the cut-off of 45 days in our combined cohort of patients, although approximately 70% patients in Survival With Oral D-Sotalol were enrolled at an average of 76 weeks after MI.

CONCLUSIONS

The use of a single dichotomy limit of LVEF to determine the need for ICD treatment leads to a trade-off between the sensitivity and positive predictive accuracy across a range of LVEF. No single dichotomy limit is completely satisfactory as all patients with an LVEF ⩽10% in our study died from non-arrhythmic causes. The cost effectiveness of ICD treatment is maximal in those with an LVEF of 16–20%.

Acknowledgments

This study was supported by a British Heart Foundation project grant (Number PG/98006). YGY was a British Heart Foundation Research Fellow in Cardiology. Professor AJC is British Heart Foundation Professor of Clinical Cardiology. We thank Dr Eric Lim for his statistical advice. YGY had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis

REFERENCES

Footnotes

  • Published Online First 18 January 2007

  • Competing interests: None.