Ability of microvolt T-wave alternans to modify risk assessment of ventricular tachyarrhythmic events: A meta-analysis

Curriculum in Cardiology

Ability of microvolt T-wave alternans to modify risk assessment of ventricular tachyarrhythmic events: A meta-analysis Anurag Gupta, MD, a,d Donald D. Hoang, BA, b,d Leah Karliner, MD, MAS, c Jeffrey A. Tice, MD, c Paul Heidenreich, MD, MS, a,b Paul J. Wang, MD, a and Mintu P. Turakhia, MD MAS a,b Stanford, Palo Alto, and San Francisco, CA

Background Prior studies have indicated that the magnitude of risk association of microvolt T-wave alternans (MTWA) testing appears to vary with the population studied. We performed a meta-analysis to determine the ability of MTWA to modify risk assessment of ventricular tachyarrhythmic events (VTEs) and sudden cardiac death (SCD) across a series of patient risk profiles using likelihood ratio (LR) testing, a measure of test performance independent of disease prevalence. Methods

We identified original research articles published from January 1990 to January 2011 that investigate spectrally derived MTWA. Ventricular tachyarrhythmic event was defined as the total and arrhythmic mortality and nonfatal sustained or implantable cardioverter-defibrillator–treated ventricular tachyarrhythmias. Summary estimates were created for positive and nonnegative MTWA results using a random-effects model and were expressed as positive (LR+) and negative (LR−) LRs.

Results

Of 1,534 articles, 20 prospective cohort studies met our inclusion criteria, consisting of 5,945 subjects predominantly with prior myocardial infarction or left ventricular dysfunction. Although there was a modest association between positive MTWA and VTE (relative risk 2.45, 1.58-3.79) and nonnegative MTWA and VTE (3.68, 2.23-6.07), test performance was poor (positive MTWA: LR+ 1.78, LR− 0.43; nonnegative MTWA: LR+ 1.38, LR− 0.56). Subgroup analyses of subjects classified as prior VTE, post–myocardial infarction, SCD-HeFT type, and MADIT-II type had a similar poor test performance. A negative MTWA result would decrease the annualized risk of VTE from 8.85% to 6.37% in MADIT-II–type patients and from 5.91% to 2.60% in SCD-HeFT–type patients.

Conclusions Despite a modest association, results of spectrally derived MTWA testing do not sufficiently modify the risk of VTE to change clinical decisions. (Am Heart J 2012;163:354-64.)

Depressed left ventricular ejection fraction (LVEF) remains the most widely used criterion for identifying patients at high risk for sudden cardiac death (SCD) and for selecting candidates for prophylactic implantable cardioverter-defibrillators (ICDs). 1-5 However, EF has a poor negative and positive predictive value for SCD, with

From the aStanford University School of Medicine, Stanford, CA, bVA Palo Alto Health Care System, Palo Alto, CA, and cDepartment of Medicine, Division of General Internal Medicine, University of California, San Francisco, San Francisco, CA. d

A.G. and D.H. are co–first authors and contributed equally to this work. Financial support: VA Health Services Research and Development Career Development Award (CDA09027-1), American Heart Association National Scientist Development Grant (09SDG2250647), and California Technology Assessment Forum (Blue Shield of California Foundation). Submitted July 18, 2011; accepted November 30, 2011. Reprint requests: Mintu P. Turakhia, MD, MAS, VA Palo Alto Health Care System, 3801 Miranda Ave 111C, Palo Alto, CA 94304. E-mail: [email protected] 0002-8703/$ - see front matter © 2012, Mosby, Inc. All rights reserved. doi:10.1016/j.ahj.2011.11.021

only 21% to 24% of patients with reduced EF receiving appropriate ICD therapies in randomized trials. 4 , 6-8 Microvolt T-wave alternans (MTWA) testing is a noninvasive test that characterizes small beat-to-beat fluctuations in T-wave morphology, amplitude, or timing. These changes may reflect temporal and spatial dispersions of ventricular repolarization, which are hypothesized mechanisms that precede ventricular tachyarrhythmias. 9 Prior studies have shown that an abnormal spectrally derived MTWA result is modestly associated with ventricular tachyarrhythmic events (VTEs) and that a normal result has a high negative predictive value. 16,17 Although risk ratios and predictive values are useful in characterizing the magnitude of association, they are highly dependent on disease prevalence. Therefore, they do not appropriately describe a test's ability to modify the risk of disease when the pretest or baseline risk is more clearly defined, such as in patients who meet the criteria for prophylactic defibrillator implantation based on the MADIT-II or SCD-HeFT trials. 10,11 We therefore

American Heart Journal Volume 163, Number 3

performed a meta-analysis to evaluate the likelihood ratio (LR)–based test performance of MTWA based on its ability to modify risk assessment of VTE across a range of patient risk profiles.

Methods Data sources and search strategy We performed a literature search of MEDLINE, Cochrane Library, International Network of Agencies for Health Technology Assessment, and Web of Science to identify original, fulllength research articles investigating spectrally derived, exercise-induced MTWA from January 1990 to January 2011. We searched using the keywords “MTWA” and “microvolt t-wave alternans” and did not limit the search by language. We identified additional potentially relevant studies by handsearching reference lists of pertinent articles and review articles. We excluded review articles, abstracts, book chapters, conference proceedings, and correspondence. Endnote software version X3.0 (Thomson Reuters, Carlsbad, CA), in conjunction with manual referencing, was used to eliminate duplicate articles. If more than one study reported on the same cohort, then we only used the most recent publication of that cohort. Articles of subcohorts were excluded if an article using the entire cohort was previously included. When the same research study was found in multiple journals or repositories, only one instance of the study was included.

Study selection We included studies that met the following criteria: (1) original, full-length published research articles investigating the diagnostic accuracy of MTWA for prediction of VTE; (2) randomized controlled trial or prospective cohort study; (3) MTWA determined by spectrally derived, predominantly exercise-induced (N75% of study population) method; (4) clear abnormal definition of MTWA provided; (5) primary data of MTWA results and VTE outcomes reported; (6) sample size of at least 30 patients; and (7) follow-up duration of at least 6 months. We included subcohorts of the total study population that met our inclusion criteria if sufficient detail was provided. We excluded studies with healthy participants or patients with Brugada or long-QT syndromes. When study numerators or denominators for MTWA results and/or VTE could not be determined for calculation of test performance estimates, we attempted to contact authors for clarification, before excluding these studies. 12-14

Data extraction Study selection and data extraction were completed independently and in duplicate by 2 reviewers (D.H. and A.G.). Disagreements were resolved by consensus with a third reviewer (M.P.T.). Microvolt T-wave alternans results were classified as negative, positive, and nonnegative (which includes positive or indeterminate results). For VTE, we accepted the outcome definitions used by the original researchers, including the following: the clinical outcomes of mortality (including arrhythmic, cardiac, and/or total mortality), nonfatal sustained ventricular tachyarrhythmias, and/or ICD-treated ventricular tachyarrhythmias. We further abstracted data regarding study design, inclusion and exclusion

Gupta et al 355

criteria, baseline patient characteristics and demographics, and follow-up duration. We evaluated the quality, design, and execution of the observational studies based on established, standardized criteria. 15-17 Because all identified studies were observational in nature and there were no clinical trials, we did not assign a quality score in the absence of a validated method for such assessment. 18-20

Data synthesis and analysis The primary outcomes of the study were the positive and negative LRs (LR+ and LR−) for normal and abnormal MTWA test results. Likelihood ratio testing, which uses a Bayesian approach that integrates sensitivity and specificity, can quantify how much the risk of a target disease (in this case, VTE/SCD) will change with a given test result (posttest probability; in this case MTWA) 21,22: Posttest odds of VTE/SCD with negative MTWA = Pretest odds of VTE/SCD × LR− Posttest odds of VTE/SCD with abnormal MTWA = Pretest odds of VTE/SCD × LR+ Generally, LR+ N5 and LR− b0.2 indicate strongly favorable use of a prognostic test. 10 Likelihood ratios can define a range of pretest probabilities for which MTWA may be useful and more appropriate for clinical decision making and are independent of the studied disease prevalence. Because some studies grouped indeterminate results along with positive results as abnormal (“non-negative MTWA”), we calculated separate sets of LRs for studies using nonnegative and positive result criteria. We used the DerSimonian and Laird random-effects method to create summary estimates for positive and nonnegative (positive and indeterminate) MTWA results using STATA version 10 (StataCorp LP, College Station, TX). 23 We performed a series of subanalyses to determine results across select patient risk profiles. Specifically, we reassessed LRs after excluding trials that investigated patients (1) with prior VTE 24-26 to characterize a primary prevention population and (2) with recent myocardial infarction (MI) within approximately 3 months. 27-30 We also assessed LRs in studies with differences of end point definitions, excluding trials in which (3) ICD analysis and/or therapies were used as a surrogate for VTE 24-26 ,31-34 and (4) surrogates for SCD (including ICD analysis and/or therapies, cardiac mortality or total mortality) were used as end points for VTE. 24-26,30,31,34-38 We further assessed LRs in trials that investigated patients with (5) a MADITII–type indication (prior MI and LVEF ≤0.30), 32,36 (6) a SCDHeFT–type indication (New York Heart Association class II or III congestive heart failure and LVEF ≤35%), 24,25,31,33-35,37,39-43 (7) prior MI 27-30,38 (8) mean LVEF N35%, 24,26,28-30,35,38,40,42 and (9) mean LVEF ≤35%. 25,27,31-34,36,37,39,41,43 We applied LRs derived from studies that evaluated MADITII–type and SCD-HeFT–type patients to their respective population's pretest probability event rate. We used SCD as the lower bound and VTE as the upper bound of risk for posttest probability of adverse events. For clinical applicability, we annualized the populations' pretest probability of SCD and VTE event rates before calculating the posttest probability. 6,44 Sudden cardiac death rates were abstracted from MADIT-II and SCD-HeFT over a mean follow-up of 20 months and a median

American Heart Journal March 2012

356 Gupta et al

Figure 1

CONSORT diagram of study selection. CONSORT diagram detailing the results of the systematic literature review including, the reasons for excluding studies from the meta-analysis.

follow-up of 45.5 months, respectively. Ventricular tachyarrhythmic event rates were calculated from the appropriate shock rate of the ICD arms of the MADIT-II and SCD-HeFT studies over a mean follow-up of 24 months and a median follow-up of 45.5 months, respectively. We used the Mantel-Haenszel method to evaluate study heterogeneity. 45 We used the Egger and Begg tests to assess

publication bias 46 and calculated Kendall τ coefficient to compare sample size with effect size. 46,47

Role of the funding source This research was partially supported by grants from the VA Health Services Research and Development Career Development Award, the American Heart Association National

American Heart Journal Volume 163, Number 3

Gupta et al 357

Table I. Characteristics of the 20 prospective studies investigating MTWA

Study

N⁎

Adachi et al 35 Baravelli et al 24 Bloomfield et al 31 Chow et al 36 Chow et al 32 Gold et al 33

64 56 549 768 575 490

Grimm et al

39

Hohnloser et al

26

Hohnloser et al 25 Ikeda et al 27 Ikeda et al 28 Ikeda et al 29 Kitamura et al 40 Klingenheben et al 41 Sakabe et al Sakabe et al

42 43

Salerno-Uriarte et al 34 Sarzi Braga et al 37 Schwab et al 38 Tapanainen et al 30 Overall

Population

NICMP ICMP or NICMP + NYHA II ICMP or NICMP + LVEF ≤40% ICMP + LVEF ≤35% ICMP + LVEF ≤30% + prior MI + ICD ICMP or NICMP + LVEF ≤35% + NYHA II/III 263 NICMP + LVEF ≤45% + LVEDd N56 mm 62 ICD recipient for VTE or NSVT with syncope 137 NICMP 102 Post-MI (mean 20 d) 739 Post-MI (mean 2.7 mo) 1003 Post-MI (mean 48 d) 83 NICMP 107 ICMP or NICMP + LVEF ≤45% + NYHA II/III + history CHF 30 NICMP 41 ICMP or NICMP + AAD for history VTE 446 NICMP + LVEF ≤40% + NYHA II/III

Mean Mean Mean LVEF (%) † age (y) † Men (%) † NYHA †

Prior VTE

Pretest Mean β-blocker follow-up stopped? (mo) †

45 36 25 27 24 24

49 65 56 67 65 59

81 86 71 82 84 76

1.6 2 1.9 NA 2 2.3

No Yes (some) No No No No

No Yes No Yes No Yes

24 17 20 18 25 30

30

48

73

NA

No

Yes

52

36

60

81

NA

Yes (all)

No

15

29 NA 51 55 41 28

55 59 NA 64 52 56

77 83 83 79 81 80

NA NA NA NA 1.7 NA

No NA NA NA No No

No No No No Yes ‡ No

14 13 25 32 21 15

32 48

55 54

90 80

1.7 1.6

NA Yes (all)

No No

13 13

30

59

8

2.2

No

No

19

44

ICMP or NICMP + CHF

29

60

89

2.4

NA

No

19

140 246

Post-MI (mean 15 d) Post-MI (mean 8.1 d)

56 45

60 62

76 72

NA 1.5

No NA

Yes No

15 14

5945

–

37

61

79

–

–

–

24

NICMP indicates Nonischemic cardiomyopathy; ICMP, ischemic cardiomyopathy; LVEDd, left ventricular end-diastolic dimension; NYHA, New York Heart Association functional classification; NSVT, nonsustained ventricular tachycardia; NA, not available; CHF, congestive heart failure; AAD, antiarrhythmic drug. ⁎ N represents the total number of patients with follow-up data (including reported VTE outcome) meeting the inclusion criteria of this meta-analysis. † Represents reported, calculated, or estimated data for the cohort of patients (N) included for analysis in this meta-analysis. ‡ Reported as “Yes” as no patients prescribed β-blocker at the time of MTWA testing.

Scientist Development Grant, and the California Technology Assessment Forum (Blue Shield of California Foundation). The funding sources had no role in the design, analysis, or reporting of the study or in the decision to submit the manuscript for publication. The authors are solely responsible for the design and conduct of this study, all study analyses, the drafting and editing of the paper, and its final contents.

Results Our literature search yielded a total of 1,534 potentially relevant articles (Figure 1). From these articles, we identified 59 studies for full-text review. Reasons for exclusion are shown in Figure 1. Twenty cohort studies with no randomized controlled trials met our inclusion criteria, representing 5,945 subjects who had follow-up data for VTE (Table I). The baseline patient characteristics of the 20 studies are shown in Table I. The mean LVEF across studies was 37%, with a range of 24% to 56%. The mean

age across studies was 61 years, and 79% of the subjects were male. Included studies exclusively comprised subjects with left ventricular dysfunction, postMI, or prior VTE. The mean follow-up across studies was 24 months, with range between studies of 13 to 52 months. The outcome definitions and study results are detailed in Table II. There were 2,272 subjects with positive MTWA results (36.8%), 2,702 with negative results (43.8%), and 1,198 with indeterminate results (19.4%). The overall event rate for VTE was 9.4%.

Test performance Summary estimates of sensitivity, specificity, positive predictive value, and negative predictive value are detailed in Figures 2 and 3. When indeterminate MTWA test results are included with positive tests as an abnormal result (nonnegative MTWA; Figure 2), the overall sensitivity was 0.79 (95% CI 0.71-0.88), the specificity was 0.45 (0.33-0.56), the positive predictive value for a

American Heart Journal March 2012

358 Gupta et al

Table II. Outcomes of each of the 20 studies Study

MTWA +

MTWA −

MTWA Ind

N⁎

30 30 162 355 293 182 137 36 66

34 26 189 254 214 135 72 26 34

18 17 † 198 159 68 173 54 26 † 37

Events

TN

FN

TP

FP

64 56 549 768 575 490 263 62 137

10 7 51 99 70 75 38 31 18

33 26 185 233 192 119 65 19 32

1 0 4 21 22 16 7 7 2

9 7 NA 44 NA 31 18 24 13

21 23 NA 311 NA 151 119 12 53

NA NA 47 78 48 59 31 NA 16

NA NA 313 436 313 296 160 NA 87

Ikeda et al 27 50 52 17 † 28 Ikeda et al 302 437 95 † Ikeda et al 29 169 747 87 Kitamura et al 40 46 37 21 † Klingenheben et al 41 52 33 22 Sakabe et al 42 24 6 8† Sakabe et al 43 30 19 15 Salerno-Uriarte 200 154 92 et al 34 Sarzi Braga et al 37 24 13 9 44 Schwab et al 38 28 76 36 140 30 Tapanainen et al 56 144 46 246 Overall 2272 (37%) 2702 (44%) 1198 (19%) 5945

15 24 18 12 13 9 21 33

51 435 744 36 33 6 12 150

1 2 3 1 0 0 5 4

14 22 15 11 11 9 16 NA

36 280 154 35 41 15 8 NA

NA NA 15 NA 13 NA NA 29

NA NA 241 NA 61 NA NA 263

7 13 0 7 16 3 74 2 1 27 2 143 1 0 56 556 (9.4%) 2601 99 252 1358

7 1 1 345

24 63 101 2358

Adachi et al 35 Baravelli et al 24 Bloomfield et al 31 Chow et al 36 Chow et al 32 Gold et al 33 Grimm et al 39 Hohnloser et al 26 Hohnloser et al 25

VTE

SCD, sVT/VF SCD, sVT/VF, ICD therapy ACM, sVT/VF, ICD therapy ACM SCD, ICD therapy SCD, sVT/VF, ICD therapy SCD, sVT/VF ICD therapy SCD, VF, unstable VT/VF (per ICD EGM) 102 sVT/VF 739 SCD, VF 1003 SCD, VF 83 SCD, sVT/VF 107 SCD, sVT/VF 30 sVT ‡ 41 sVT/VF 446 Cardiac death, sVT/VF Cardiac death ACM, sVT/VF ACM –

TNN FNN

MTWA + indicates Positive MTWA result; MTWA −, negative MTWA result; MTWA Ind, indeterminate MTWA result; TN, true-negative MTWA results; FN, false-negative MTWA results; TP, true-positive MTWA results; FP, false-positive MTWA results; TNN, true-nonnegative (ie, includes positive and indeterminate) MTWA results; FNN, false-nonnegative (ie, includes positive and indeterminate) MTWA results; sVT, sustained VT; VF, ventricular fibrillation; NA, not available; ACM, all-cause mortality; EGM, electrogram. ⁎ N represents the total number of patients with follow-up data, including reported VTE outcome, who met the inclusion criteria of this meta-analysis and were included for analysis. † Study authors did not subsequently include for analysis and/or report data from patients with indeterminate MTWA results. ‡ Authors further reported NSVT outcomes in their study; NSVT not included as VTE end point in this study.

nonnegative MTWA result was 0.11 (0.08-0.15), the negative predictive value was 0.95 (0.92-0.98), and the relative risk was 2.45 (1.58-3.79). When indeterminate MTWA test results are excluded from the study cohorts so that a positive result is compared with a negative result (Figure 3), the results were similar: the overall sensitivity was 0.80 (0.73-0.88), the specificity was 0.60 (0.51-0.68), the positive predictive value was 0.13 (0.08-0.19), the negative predictive value was 0.96 (0.93-0.99), and the relative risk was 3.68 (2.23-6.07). Test performance, expressed in positive and negative LRs, including paired forest plots, are also shown in Figures 2 and 3. Overall, test performance was poor for positive and nonnegative MTWA results. For nonnegative MTWA result vs negative MTWA result (Figure 2), LR+ was 1.38 (1.18-1.62). For positive vs negative MTWA (indeterminates excluded), LR+ was 1.78 (1.43-2.22). No studies had an LR+ of ≥5, and only 1 had an LR+ N3. 29 Test performance for negative results was modestly better. For nonnegative MTWA result vs negative MTWA results, LR− was 0.56 (0.40-0.77). For positive vs negative MTWA (indeterminates excluded; Figure 3), LR− was 0.43 (0.30-0.63). Only 2 studies comparing nonnegatives to negatives had an LR− of b0.2. 37,41 Among 17 studies that excluded indeterminate results, LR− was b0.2 in 8 studies.

Subgroup analysis of 9 subpopulations Subgroup analysis on 9 subpopulations is detailed in Table III. These subgroups are divided based on study end points and enrollment criteria. When including indeterminate MTWA results, the range of positive LR for nonnegative MTWA results was 1.18 to 1.79 for each of the subanalyses, which was comparable with the overall summary estimate of 1.38 (1.18-1.62). Likewise, the range of negative LRs of 0.34 to 0.72 for each of the subanalyses was similar to the overall summary estimate of 0.56 (0.400.77). Likelihood ratio estimates were best for studies including only outcome of SCD and excluding studies with other surrogate outcomes (Table III). Negative likelihood ratio estimates for studies of MADIT-II, SCDHeFT, LVEF N35%, and LVEF ≤35% were poor overall and not better than the overall results. The trends were similar when excluding indeterminate MTWA test results (data not shown). Patients with indications for prophylactic ICD implantation In studies evaluating MADIT-II–type and SCD-HeFT– type patients, the magnitude of risk modification from pretest to posttest probabilities with MTWA results was small. In MADIT-II–type patients, the annualized risk of

American Heart Journal Volume 163, Number 3

Figure 2

Paired forest plots of LR+ and LR− (including indeterminate MTWA results; ie, comparing nonnegative [positive and indeterminate] to negative MTWA results).

Gupta et al 359

360 Gupta et al

Figure 3

Paired forest plot of LR+ and LR− (excluding indeterminate MTWA results; ie, comparing positive to negative MTWA results). American Heart Journal March 2012

American Heart Journal Volume 163, Number 3

Gupta et al 361

Table III. Likelihood ratios of 9 subpopulations LR+ Overall Excluding studies with outcome of ICD therapies⁎ Excluding studies with outcome of ICD therapies, total mortality, or cardiac mortality § Excluding studies of subjects with prior VTE † Excluding studies of subjects with recent MI ‡ Studies of MADIT-II–type subjects || Studies of SCD-HeFT–type subjects ¶ Studies of subjects post-MI # Subjects of subjects with LVEF N35%⁎⁎ Studies of subjects with LVEF ≤35% ††

1.38 1.55 1.79 1.42 1.29 1.18 1.31 1.65 1.65 1.27

(1.18-1.62) (1.14-2.11) (0.99-3.24) (1.20-1.69) (1.17-1.42) (1.07-1.30) (1.17-1.46) (0.43-6.29) (0.43-6.29) (1.16-1.39)

LR− 0.56 0.57 0.34 0.56 0.56 0.72 0.44 0.62 0.62 0.56

(0.40-0.77) (0.35-0.94) (0.12-0.95) (0.40-0.79) (0.41-0.77) (0.53-0.97) (0.27-0.72) (0.15-2.57) (0.15-2.57) (0.41-0.77)

n

Events

4768 2571 2438 4631 3519 1343 2036 1389 1389 3379

556 271 295 479 497 169 294 62 128 428

Summary estimates, reported and LR+ and LR− (when indeterminate MTWA test results were included), of the entire 20 trials included in the meta-analysis as well as 3 subpopulations. Studies included in each subanalysis are cited. ⁎ Excludes studies including ICD analysis and/or therapies as a surrogate for VTE.29,30,36-39,41 † Excludes studies including patients with prior VTE, that is, patients with aborted SCD.29-34,36-39,41 ‡ Excludes studies predominantly enrolling patients with MI within approximately 3 months of TWA testing.25,31-34,36-39,41 § Excludes studies including ICD analysis and/or therapies, total mortality, or cardiac mortality as surrogates for VTE.29,32,33,39,41 || Includes studies with MADIT-2–type subjects (the MADIT-2 trial included individuals with a prior MI and LVEF ≤30%).32,36 ¶ Includes studies with SCD-HeFT–type subjects (the SCD-HeFT trial included individuals with New York Heart Association class II or III congestive heart failure and LVEF ≤35%).25,31,33,34,37,39,41 # Includes studies with prior or recent MI.29,30,38 ⁎⁎ Includes studies in which overall mean LVEF of patients N35%.29,30,38 †† Includes studies in which overall mean LVEF of patients ≤35%.25,31-34,36,37,39,41

SCD was 6.00%. 7 Use of MTWA testing would divide subjects into high-risk group (positive or indeterminate MTWA result) with a 7.1% risk of SCD and low-risk group (negative MTWA) with a 4.3% risk (Table IV), with an absolute risk difference in reclassification of 2.8%. Similarly, in SCD-HeFT–type patients, the annualized risk of SCD was 2.95%. 8 Microvolt T-wave alternans results would divide subjects into a high-risk group with a 3.9% risk of SCD and a low-risk group with a 1.3% risk, with an absolute risk difference in reclassification of 2.6%. The results did not change when the end point of VTE, rather than SCD, was used to define pretest and posttest probabilities. Results for VTE event rates indicated an absolute risk difference in reclassification of 4.1% for MADIT-II–type patients and a difference of 5.1% for SCD-HeFT–type patients (Table IV).

Statistical heterogeneity and publication bias There was no statistical heterogeneity per the MantelHaenszel method (P = 0.295). Using the Begg test and Egger test, 47 we found no evidence of publication bias (P = .66 and P = .14, respectively).

Discussion We found that although the MTWA had a reasonable negative predictive value for the outcomes of VTE, the results of spectrally derived MTWA testing do not sufficiently modify the risk assessment of VTE or SCD to affect clinical decision making. At best, posttest probability of VTE could be reduced to 6.4% per year for MADIT-II–type patients and 2.5% per year for SCD-HeFT– type patients. Therefore, the modification of risk by a

negative MTWA result would be unlikely to change the decision to consider prophylactic ICD implantation in patients with a reduced EF, unless the treatment threshold is above these posttest probabilities. These findings were consistent across summary estimates for the entire meta-analysis, which analyzed MTWA test performance in all patients at elevated risk for SCD, and in sensitivity analyses of 9 specific subpopulations (Table II). Previous smaller and older meta-analyses have found similar negative predictive values. Gehi and colleagues 48 performed a meta-analysis of 19 studies published through 2004, consisting of 2,608 individuals over a mean follow-up of 21 months (including 17 trials that did not use ICD therapies as an end point). They reported summary estimates (excluding indeterminate MTWA findings) of the following: relative risk, 3.77 (2.39-5.95); positive predictive value, 0.193 (0.177-0.210); negative predictive value, 0.972 (0.965-0.979); however, they did not assess pretest risk modification. A meta-analysis by van der Avoort and colleagues 49 analyzed 9 prospective studies of 1,946 patients and reported a relative risk of 2.7 (1.4-6.1) for MTWA (excluding indeterminate results) and concluded that the clinical value of MTWA was less clear. A subsequent review by Myles and colleagues 50 concluded that the benefit of MTWA seemed to vary with the population studied, and further data were required before its incremental prognostic role could be properly defined. We now address this issue in our meta-analysis. By evaluating LRs, which are independent of disease prevalence, we found that the role of MTWA in clinical decision making is limited based on the current data, particularly in patients whose risk is already elevated, such as patients with reduced EF. Furthermore, we also

American Heart Journal March 2012

362 Gupta et al

Table IV. Likelihood ratios applied to MADIT-II and SCD-HeFT populations Annualized pretest probability of SCD MADIT-II, 7 6.00%

×

LR of VTE

=

Annualized posttest Probability of SCD

Including indeterminate results⁎

1.18 (LR+) 0.72 (LR−) 1.18 (LR+) 0.75 (LR−) 1.31 (LR+) 0.44 (LR−) 1.56 (LR+) 0.42 (LR−)

=

7.1% 4.3% 7.1% 4.5% 3.9% 1.3% 4.6% 1.2%

Excluding indeterminate results⁎ SCD-HeFT, 8 2.95%

Including indeterminate results⁎ Excluding indeterminate results⁎

Annualized pretest probability of VTEs MADIT-II, 7 8.85%

= =

×

LR of VTE

=

Annualized posttest Probability of VTE

Including indeterminate results⁎

1.18 (LR+) 0.72 (LR−) 1.18 (LR+) 0.75 (LR−) 1.31 (LR+) 0.44 (LR−) 1.56 (LR+) 0.42 (LR−)

=

10.4% 6.4% 10.4% 6.6% 7.7% 2.6% 9.2% 2.5%

Excluding indeterminate results⁎ SCD-HeFT, 8 5.91%

=

Including indeterminate results⁎ Excluding indeterminate results⁎

= = =

These calculations demonstrate the modification of pretest probability of SCD and VTE based on MTWA results in MADIT-II–type and SCD-HEFT–type patients. We applied LRs to these populations with SCD as the lower bound of risk and VTE as the upper bound. The absolute risk difference in reclassification is computed as the difference between the posttest probabilities of the high- and low-risk groups. Based on an annualized pretest risk of SCD of 6.0% in MADIT-II–type patients, MTWA results would increase the absolute risk by 1.08% (if positive) and decrease the absolute risk by 1.68% (if negative). For SCD-HeFT–type patients with an annualized pretest risk of 2.95%, MTWA testing would increase the risk by 0.91% or decrease the risk by 1.65%. The risk differences are similar when indeterminate results are excluded and are also similar when LRs are applied to VTE pretest probabilities. For MTWA testing to affect clinical decisions, the treatment threshold would have to be between SCD annual risks of 4.32% to 7.08% and 1.30% to 3.86%, and between VTE annual risks of 6.37% to 10.44% and 2.60% to 7.74% in MADIT-II–type and SCD-HeFT–type patients, respectively. ⁎ “Including indeterminate results” refers to LRs derived from comparing nonnegative MTWA results (ie, positive and indeterminate results) to negative MTWA results. “Excluding indeterminate results” refers to LRs derived from comparing positive MTWA results to negative MTWA results.

derived point estimates with and without studies using indeterminate MTWA results and found that LRs and summary estimates are not largely influenced by inclusion or exclusion of indeterminate test results. The included 20 studies use a number of different end points to evaluate MTWA performance as opposed to the sole end point of SCD. No trial used SCD as the singular end point, and the VTE end points used in many of the individual trials included in this meta-analysis potentially included nonarrhythmic or other surrogates such as ICD therapies, cardiac mortality, and total mortality. Hohnloser and colleagues 51 also attempted to assess the potential influence of surrogate end points on MTWA's performance. They reported a significantly higher hazard ratio (HR) for a nonnegative MTWA when comparing trials enrolling few patients with ICD (HR 13.6 [8.5-30.4]) to trials enrolling a higher proportion of patients with ICD (HR 1.6 [1.2-2.1]). 51 In contrast to the study of Hohnloser et al, our analysis reviewed the literature through January 2011 (as opposed to November 2007) and applied stricter inclusion criteria (including only full-length published articles). We also did not convert event data into annual event rates for the analysis because we believed that there was insufficient data on survival curves to calculate these values. Rather, we derived LRs and generated posttest probabilities from annualized MADIT-II and SCD-

HeFT event rates, which are more stable because of the large sample size and longer follow-up of these trials. Still, both our study and the study of Hohnloser et al included studies with surrogate end points beyond SCD to evaluate MTWA test performance. Future, carefully designed, prospective studies with well-defined populations and end points could help clarify the role of MTWA. Notwithstanding, the limitations posed by surrogate end points highlight the difficulties in evaluating MTWA and the shortcomings of prior MTWA research with respect to clinical decision making. There are several limitations to our meta-analysis. First, all data were derived from prospective cohort studies of spectral MTWA. We did not identify any randomized trials meeting our inclusion criteria. Second, the 20 included studies differed with respect to study design, patient characteristics, timing of MTWA testing, VTE outcomes, and follow-up duration (Tables I and II). Nevertheless, there was no statistical evidence of heterogeneity. Moreover, our overall meta-analysis and subanalyses, which characterized more similar populations, all yielded comparable summary estimates for MTWA (Table III). Finally, more recent studies have shown that MTWA used in combination with invasive electrophysiology study or heart rate recovery may have better risk prediction. 13,52,53 Further research is required to determine if

American Heart Journal Volume 163, Number 3

multivariable or sequential risk stratification could be clinically useful.

Conclusion Despite a modest association with VTE, the results of spectrally derived MTWA testing do not sufficiently modify risk assessment of VTE to change clinical decisions, particularly if conventional risk markers such as reduced EF indicate an increased pretest risk. Based on current evidence, routine incorporation of MTWA results into clinical decision making regarding ICD therapy is not warranted.

Disclosures Disclaimer: The content is solely the responsibility of the authors and does not necessarily represent the official views of the Department of Veterans Affairs. Conflicts of interest: None.

References 1. Bardy GH, Lee KL, Mark DB, et al. Amiodarone or an implantable cardioverter-defibrillator for congestive heart failure. N Engl J Med 2005;352:225-37. 2. Buxton AE, Lee KL, Fisher JD, et al. randomized study of the prevention of sudden death in patients with coronary artery disease. Multicenter Unsustained Tachycardia Trial Investigators. N Engl J Med 1999;341:1882-90. 3. Epstein AE, DiMarco JP, Ellenbogen KA, et al. ACC/AHA/HRS 2008 Guidelines for Device-Based Therapy of Cardiac Rhythm Abnormalities: a report of the American College of Cardiology/American Heart Association Task Force on Practice Guidelines (Writing Committee to revise the ACC/AHA/NASPE 2002 Guideline Update for Implantation of Cardiac Pacemakers and Antiarrhythmia Devices): developed in collaboration with the American Association for Thoracic Surgery and Society of Thoracic Surgeons. Circulation 2008;117: e350-408. 4. Moss AJ, Greenberg H, Case RB, et al. Long-term clinical course of patients after termination of ventricular tachyarrhythmia by an implanted defibrillator. Circulation 2004;110:3760-5. 5. Moss AJ, Zareba W, Hall WJ, et al. Prophylactic implantation of a defibrillator in patients with myocardial infarction and reduced ejection fraction. N Engl J Med 2002;346:877-83. 6. Daubert JP, Zareba W, Cannom DS, et al. Inappropriate implantable cardioverter-defibrillator shocks in MADIT II: frequency, mechanisms, predictors, and survival impact. J Am Coll Cardiol 2008;51: 1357-65. 7. Greenberg H, Case RB, Moss AJ, et al. Analysis of mortality events in the Multicenter Automatic Defibrillator Implantation Trial (MADIT-II). J Am Coll Cardiol 2004;43:1459-65. 8. Packer DL, Prutkin JM, Hellkamp AS, et al. Impact of implantable cardioverter-defibrillator, amiodarone, and placebo on the mode of death in stable patients with heart failure: analysis from the sudden cardiac death in heart failure trial. Circulation 2009;120:2170-6. 9. Narayan SM. T-wave alternans and the susceptibility to ventricular arrhythmias. J Am Coll Cardiol 2006;47:269-81. 10. Deeks JJ. Systematic reviews in health care: systematic reviews of evaluations of diagnostic and screening tests. BMJ 2001;323: 157-62.

Gupta et al 363

11. Pepe MS, Janes H, Longton G, et al. Limitations of the odds ratio in gauging the performance of a diagnostic, prognostic, or screening marker. Am J Epidemiol 2004;159:882-90. 12. Cantillon DJ, Stein KM, Markowitz SM, et al. Predictive value of microvolt T-wave alternans in patients with left ventricular dysfunction. J Am Coll Cardiol 2007;50:166-73. 13. Costantini O, Hohnloser SH, Kirk MM, et al. The ABCD (Alternans Before Cardioverter Defibrillator) Trial: strategies using T-wave alternans to improve efficiency of sudden cardiac death prevention. J Am Coll Cardiol 2009;53:471-9. 14. Rashba EJ, Osman AF, Macmurdy K, et al. Enhanced detection of arrhythmia vulnerability using T wave alternans, left ventricular ejection fraction, and programmed ventricular stimulation: a prospective study in subjects with chronic ischemic heart disease. J Cardiovasc Electrophysiol 2004;15:170-6. 15. Downs SH, Black N. The feasibility of creating a checklist for the assessment of the methodological quality both of randomised and non-randomised studies of health care interventions. J Epidemiol Community Health 1998;52:377-84. 16. von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. J Clin Epidemiol 2008;61:344-9. 17. Whiting P, Rutjes AW, Reitsma JB, et al. The development of QUADAS: a tool for the quality assessment of studies of diagnostic accuracy included in systematic reviews. BMC Med Res Methodol 2003;3:25. 18. Altman DG. Systematic reviews of evaluations of prognostic variables. BMJ 2001;323:224-8. 19. Juni P, Altman DG, Egger M. Systematic reviews in health care: assessing the quality of controlled clinical trials. BMJ 2001;323:42-6. 20. Juni P, Witschi A, Bloch R, et al. The hazards of scoring the quality of clinical trials for meta-analysis. JAMA 1999;282:1054-60. 21. Irwig L, Tosteson AN, Gatsonis C, et al. Guidelines for meta-analyses evaluating diagnostic tests. Ann Intern Med 1994;120:667-76. 22. Jaeschke R, Guyatt GH, Sackett DL. Users' guides to the medical literature. III. How to use an article about a diagnostic test. B. What are the results and will they help me in caring for my patients? The evidence-based Medicine Working Group. JAMA 1994;271:703-7. 23. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials 1986;7:177-88. 24. Baravelli M, Salerno-Uriarte D, Guzzetti D, et al. Predictive significance for sudden death of microvolt-level T wave alternans in New York Heart Association class II congestive heart failure patients: a prospective study. Int J Cardiol 2005;105:53-7. 25. Hohnloser SH, Klingenheben T, Bloomfield D, et al. Usefulness of microvolt T-wave alternans for prediction of ventricular tachyarrhythmic events in patients with dilated cardiomyopathy: results from a prospective observational study. J Am Coll Cardiol 2003;41: 2220-4. 26. Hohnloser SH, Klingenheben T, Li YG, et al. T wave alternans as a predictor of recurrent ventricular tachyarrhythmias in ICD recipients: prospective comparison with conventional risk markers. J Cardiovasc Electrophysiol 1998;9:1258-68. 27. Ikeda T, Kumagai K, Takami M, et al. Lack of an association between T-wave alternans and ST-segment depression during exerciseinduced ambulatory ischemia. J Electrocardiol 2000;33:261-7. 28. Ikeda T, Saito H, Tanno K, et al. T-wave alternans as a predictor for sudden cardiac death after myocardial infarction. Am J Cardiol 2002;89:79-82. 29. Ikeda T, Yoshino H, Sugi K, et al. Predictive value of microvolt T-wave alternans for sudden cardiac death in patients with preserved cardiac

American Heart Journal March 2012

364 Gupta et al

30.

31.

32.

33.

34.

35.

36.

37.

38.

39.

40.

function after acute myocardial infarction: results of a collaborative cohort study. J Am Coll Cardiol 2006;48:2268-74. Tapanainen JM, Still AM, Airaksinen KE, et al. Prognostic significance of risk stratifiers of mortality, including T wave alternans, after acute myocardial infarction: results of a prospective follow-up study. J Cardiovasc Electrophysiol 2001;12:645-52. Bloomfield DM, Bigger JT, Steinman RC, et al. Microvolt T-wave alternans and the risk of death or sustained ventricular arrhythmias in patients with left ventricular dysfunction. J Am Coll Cardiol 2006;47: 456-63. Chow T, Kereiakes DJ, Onufer J, et al. Does microvolt T-wave alternans testing predict ventricular tachyarrhythmias in patients with ischemic cardiomyopathy and prophylactic defibrillators? The MASTER (Microvolt T Wave Alternans Testing for Risk Stratification of Post-Myocardial Infarction Patients) trial. J Am Coll Cardiol 2008;52: 1607-15. Gold MR, Ip JH, Costantini O, et al. Role of microvolt T-wave alternans in assessment of arrhythmia vulnerability among patients with heart failure and systolic dysfunction: primary results from the T-wave alternans sudden cardiac death in heart failure trial substudy. Circulation 2008;118:2022-8. Salerno-Uriarte JA, De Ferrari GM, Klersy C, et al. Prognostic value of T-wave alternans in patients with heart failure due to nonischemic cardiomyopathy: results of the ALPHA Study. J Am Coll Cardiol 2007;50:1896-904. Adachi K, Ohnishi Y, Yokoyama M. Risk stratification for sudden cardiac death in dilated cardiomyopathy using microvolt-level T-wave alternans. Jpn Circ J 2001;65:76-80. Chow T, Kereiakes DJ, Bartone C, et al. Prognostic utility of microvolt T-wave alternans in risk stratification of patients with ischemic cardiomyopathy. J Am Coll Cardiol 2006;47:1820-7. Sarzi Braga S, Vaninetti R, Laporta A, et al. T wave alternans is a predictor of death in patients with congestive heart failure. Int J Cardiol 2004;93:31-8. Schwab JO, Weber S, Schmitt H, et al. Incidence of T wave alternation after acute myocardial infarction and correlation with other prognostic parameters: results of a prospective study. PacePacing Clin Electrophysiol 2001;24:957-61. Grimm W, Christ M, Bach J, et al. Noninvasive arrhythmia risk stratification in idiopathic dilated cardiomyopathy: results of the Marburg Cardiomyopathy Study. Circulation 2003;108:2883-91. Kitamura H, Ohnishi Y, Okajima K, et al. Onset heart rate of microvolt-level T-wave alternans provides clinical and prognostic

41.

42.

43.

44.

45. 46. 47. 48.

49.

50.

51. 52.

53.

value in nonischemic dilated cardiomyopathy. J Am Coll Cardiol 2002;39:295-300. Klingenheben T, Zabel M, D'Agostino RB, et al. Predictive value of T-wave alternans for arrhythmic events in patients with congestive heart failure. Lancet 2000;356:651-2. Sakabe K, Ikeda T, Sakata T, et al. Predicting the recurrence of ventricular tachyarrhythmias from T-wave alternans assessed on antiarrhythmic pharmacotherapy: a prospective study in patients with dilated cardiomyopathy. Ann Noninvasive Electrocardiol 2001;6:203-8. Sakabe K, Ikeda T, Sakata T, et al. Comparison of T-wave alternans and QT interval dispersion to predict ventricular tachyarrhythmia in patients with dilated cardiomyopathy and without antiarrhythmic drugs: a prospective study. Jpn Heart J 2001;42:451-7. Poole JE, Johnson GW, Hellkamp AS, et al. Prognostic importance of defibrillator shocks in patients with heart failure. N Engl J Med 2008; 359:1009-17. Mantel N, Brown C, Byar DP. Tests for homogeneity of effect in an epidemiologic investigation. Am J Epidemiol 1977;106:125-9. Begg CB, Mazumdar M. Operating characteristics of a rank correlation test for publication bias. Biometrics 1994;50:1088-101. Egger M, Smith GD, Schneider M, et al. Bias in meta-analysis detected by a simple, graphical test. BMJ 1997;315:629-34. Gehi AK, Stein RH, Metz LD, et al. Microvolt T-wave alternans for the risk stratification of ventricular tachyarrhythmic events—a meta-analysis. JAm Coll Cardiol 2005;46:75-82. van der Avoort CJ, Filion KB, Dendukuri N, et al. Microvolt T-wave alternans as a predictor of mortality and severe arrhythmias in patients with left-ventricular dysfunction: a systematic review and meta-analysis. BMC Cardiovasc Disord 2009:9. Myles RC, Jackson CE, Tsorlalis I, et al. Is microvolt T-wave alternans the answer to risk stratification in heart failure? Circulation 2007; 116:2984-91. Hohnloser SH, Ikeda T, Cohen RJ. Evidence regarding clinical use of microvolt T-wave alternans. Heart Rhythm 2009;6(3 Suppl):S36-44. Amit G, Rosenbaum DS, Super DM, et al. Microvolt T-wave alternans and electrophysiologic testing predict distinct arrhythmia substrates: Implications for identifying patients at risk for sudden cardiac death. Heart Rhythm 2010;7:763-8. Leino J, Minkkinen M, Nieminen T, et al. Combined assessment of heart rate recovery and T-wave alternans during routine exercise testing improves prediction of total and cardiovascular mortality: the Finnish Cardiovascular Study. Heart Rhythm 2009;6:1765-71.