Microvolt T-wave Alternans

I am writing to comment on CMS’s proposed decision to exclude the Modified Moving Average (MMA) method from coverage. My training and experience are in signal processing. In collaboration with Dr. Verrier, I developed an early spectral technique to measure T-wave alternans, and we published our study in Science in 1991 (1). We selected the spectral method of “complex demodulation” approach over the spectral Fast Fourier Transform (FFT) method because it is inherently more dynamic and robust, making it capable of detecting TWA during rapidly changing physiologic conditions including myocardial ischemia and reperfusion. The capacity to track cardiac electrical instability accurately was shown not only in animals but in humans undergoing angioplasty (2).

We soon realized that even the robust method of complex demodulation was susceptible to noise and movement artifact. For this reason, we developed the Modified Moving Average approach to TWA analysis, which builds on the principle that recursive averaging is able to reject random noise that interferes with detection of the ECG signal (3). The MMA also proved to be extremely dynamic in its capacity to quantify rapidly changing signals and thus is suitable for measurement of TWA not only during routine exercise testing (4) but also during ambulatory ECG monitoring (5,6). Extensive testing with simulated ECGs and investigations of ECG signals in animals and humans showed the high accuracy of this technique. In fact, the correlation between MMA and the Fast Fourier Transform spectral method is in excess of 0.9. The main difference is that because TWA is asymmetrical and MMA seeks a peak value, versus a mean value with SM, there is an inherent scale factor of 4 to 5.

Most critical is the fact that the two methods are linearly correlated and measure the same phenomenon. The spectral method performs the analysis in the frequency domain as an ABAB pattern that repeats itself on the basis of every other beat (0.5 cycle/beat), whereas MMA detects the ABAB repeating pattern as a dichotomous beat stream. Importantly, the time-domain approach preserves the ECG morphology and enables generation of visualizable ABAB patterns in templates that distinguish A from B forms to a resolution of 20 microvolts. We showed this pattern in patients with myocardial infarction who subsequently experienced sudden cardiac death (Fig. 1 in reference 5). Recently, Dr. Stein has validated our clinical investigation and provided further examples of computer-aided visualization of the ABAB pattern in MMA templates of post-MI patients with left ventricular dysfunction (6). The use of templates with QRS-aligned superimposed beats is described in detail in GE Healthcare’s Physicians’ Guide. Also, it should be noted that our original software has been further improved in GE’s commercialized software with the addition of sophisticated baseline wander and noise reduction filters, which draw upon their decades of experience in handling noisy ECG signals that are encountered in real-life clinical practice during routine exercise testing and ambulatory ECG monitoring.

The key point is that the MMA and SM methods are measuring the same phenomenon, namely TWA, with an equivalent FDA-cleared resolution of 1 microvolt and the output of the two methods is highly correlated. Clinical studies show that the two methods also provide similar odds ratios for predicting sudden cardiac death or cardiac arrest (4-7). Therefore, it is puzzling to me why the requirement is being imposed that the MMA studies should largely duplicate the spectral FFT method literature, since the same phenomenon is being measured with equivalent accuracy. It is as if new blood pressure technology with improved acoustical detection that circumvents the need for a blood pressure cuff would be required to repeat clinical studies of the importance of hypertension in the occurrence of cardiac events.

There are also significant cost-savings with the MMA approach. Because this method achieves accurate detection of TWA with signal processing, it eliminates the need for expensive consumable supplies, namely the specialized electrodes used with the spectral FFT technique. Each set of these nonstandard electrodes costs more than $100, adding considerably to the expense of each test. Also, the MMA permits assessment of TWA in the typical flow of clinical evaluation during routine exercise and Holter monitoring, avoiding the need to schedule a separate, nonstandard test.

There is ample evidence that TWA testing can greatly reduce the cost associated with unnecessary ICD implantation, reducing the proportion of implants required to save one life from 18:1 to 7:1. MMA has been shown to be equivalent to the spectral FFT method in the REFINE study (7) and its predictive accuracy has been demonstrated in the FINCAVAS (4), ATRAMI (5), and EPHESUS (6) studies. In my opinion, the current decision to exclude MMA from coverage deprives CMS patients of a valuable diagnostic test and would be a lost opportunity to reduce excess healthcare costs.

Sincerely,

Bruce D. Nearing, Ph.D.
Assistant Professor of Medicine, Harvard Medical School
Department of Medicine, Beth Israel Deaconess Medical Center

REFERENCES
1. Nearing BD, Huang AH, Verrier RL. Dynamic tracking of cardiac vulnerability by complex demodulation of the T-wave. Science 1991;252:437-40.
2. Nearing BD, Oesterle SN, Verrier RL. Quantification of ischaemia-induced vulnerability by precordial T-wave alternans analysis in dog and human. Cardiovasc Res 1994:28:1440-9.
3. Nearing BD, Verrier RL. Modified moving average method for T-wave alternans analysis with high accuracy to predict ventricular fibrillation. J Appl Physiol 2002;92:541-549.
4. Nieminen T, Lehtimäki T, Viik J, Lehtinen R, Nikus K, Kööbi T, Niemelä K, Turjanmaa V, Kaiser W, Huhtala H, Verrier RL, Huikuri H, Kähönen M. T-wave alternans predicts mortality in a population undergoing a clinically indicated exercise test. Eur Heart J 2007; 28:2332-2337.
5. Verrier RL, Nearing BD, LaRovere MT, Pinna GD, Mittleman MA, Bigger JT, Schwartz PJ for the ATRAMI Investigators. Ambulatory ECG-based tracking of T-wave alternans in post-myocardial infarction patients to assess risk of cardiac arrest or arrhythmic death. J Cardiovasc Electrophysiol 2003;14:705-711.
6. Stein PK, Sanghavi D, Domitrovich PP, Deedwania P. Ambulatory ECG-based T-wave alternans predicts sudden cardiac death in high-risk post-MI patients with left ventricular dysfunction in the EPHESUS Study. J Cardiovasc Electrophysiol, in press.
7. Exner DV, Kavanagh KM, Slawnych MP, Mitchell LB, Ramadan D, Aggarwal SG, Noullett C, Van Schaik A, Mitchell RT, Shibata MA, Gulamhussein S, McMeekin J, Tymchak W, Schnell G, Gillis AM, Sheldon RS, Fick GH, Duff HJ for the REFINE Investigators. Noninvasive Risk Assessment Early After a Myocardial Infarction. The Risk Estimation Following Infarction, Noninvasive Evaluation (REFINE) Study. J Am Coll Cardiol 2007; 50:2275-2284.

Less

March 15, 2008

Dear Dr. Phurrough:

The American College of Cardiology (ACC; the College) appreciates the opportunity to comment on the Centers for Medicare and Medicaid Services’ Proposed Decision Memorandum for Microvolt T-wave Alternans (CAG-00293R).

The ACC is a 34,000 member non-profit professional medical society and teaching institution whose mission is to advocate for quality cardiovascular care through education, research promotion, development and application of standards and guidelines, and to influence health care policy. Our goal is to assist CMS in making appropriate coverage decisions based on scientific evidence.

The ACC agrees with CMS’ proposed National Coverage Decision (NCD) memorandum denying expansion of coverage for Microvolt T-wave Alternans (MTWA) using the Modified Moving Average (MMA) method for evaluating beneficiaries at risk of Sudden Cardiac Death (SCD) at this time. The College also concurs with CMS’ determination that there is insufficient clinical evidence on hand to support extending coverage for use of the MMA method, and that additional research is needed before such an expansion in coverage is warranted.

We appreciate your attention to this letter, and remain eager to assist you and your staff as it finalizes this NCD. If you have any questions, please contact Sergio A. Santiviago, Senior Specialist, Regulatory Affairs at 202.375.6392, or by e-mail at ssantivi@acc.org.

Sincerely,

Less

I would like to speak to the comparative value of the MMA and spectral methods for TWA measurement from the perspective of a cardiologist in an active private practice in Florida. I have used the MMA method during routine exercise stress testing for several years. Although I do not have direct experience with the spectral method, what I gather from my colleagues is that they do not find it as useful as I find the MMA. If the MMA is markedly positive, I have elected to treat with beta-blockers, even when the electrophysiology study failed to induce the VT seen on the Holter. TWA is not at all like being pregnant - either you are, or you''re not. The broad range displayed by the MMA, to me, is much more useful than either a "positive, negative, or non-negative" finding. Patients will return for follow up alternans testing to observe the effects of beta-blockers. Where the MMA may really achieve benefit is in patients who are more in a gray zone of alternans 45 to 65 microvolts, as recommended in the 2007 Nieminen et al publication in European Heart Journal. If medical treatment can reduce the alternans, then the patient is at least getting optimized medical care, and progress can be subsequently measured. Had the test been spectral and the result was non-negative, that would not necessarily encourage treatment and monitoring progress would be quite difficult. I believe that my experience with TWA testing in my cardiac patients is directly relevant to the requirement articulated in the CMS analysis section VIII, with reference to 42 CFR 410.32(a), which states in part, “…diagnostic tests must be ordered by the physician who is treating the beneficiary… for a specific medical problem and who uses the results in the management of the beneficiary’s specific medical problem.” I also wish to indicate that I find that the body of science behind the MMA method is impressive and is published in leading journals such as Circulation and Circulation Research. Furthermore, the recent studies using this method are sizeable, with the Nieminen study having over 1000 patients. As I mentioned above, this study addresses an important patient group that is very relevant to my practice. I also find that the information provided in the study by Exner and coworkers (JACC 2007) is extremely valuable and convincing, as again in a large study, this group showed in their side-by-side comparison that the MMA and spectral methods have comparable prediction. This fact is especially important as the MMA is much more compatible with the routine flow of clinical evaluation, since it can be used during routine exercise testing and does not require costly specialized electrodes. Based on my personal experience and sound studies of MMA-based TWA testing, I think it would be unfortunate if this valuable test were not reimbursed. The net result of non-coverage would be to deprive the CMS patient population of this valuable test that could save many lives and help to minimize the excess burden of ICD implantation.

Less

I am writing related to the recent CMS decision onreimbursement for repolarization alternanstesting. Thank you for the opportunity to comment.As a clinician and researcher who has used andstudied both the spectral method and the modifiedmoving average (MMA) method I feel that I canprovide useful insight to assist in your decisionmaking. I am also the Principal Investigator forthe Risk Evaluation Following Infraction,Noninvasive Evaluate or REFINE study, a trial thatcompared a variety of noninvasive tests head tohead, including the spectral and MMA methods. (JAm Coll Cardiol 2007;50:2275-84)

REFINE Study

As a brief background, REFINE was a prospectivemulticenter cohort trial supported by the CanadianGovernment and other Canadian scientific fundingagencies. Unrestricted equipment donations werereceived from both Cambridge Heart and GEHealthcare. Comparison of spectral and MMAalternans was a pre-specified comparison(http://clinicaltrials.gov/ NCT00399503).

A total of 322 patients were enrolled from sixCanadian hospitals. Patients with ejectionfraction values < 0.50 in the initial 48 hoursafter myocardial infarction (MI) were eligible. Patients were assessed in the acute (initialmonth) and non-acute (3 months) post-MI periodsand followed on average for four years. Theprimary pre-specified outcome was cardiac death orresuscitated cardiac arrest. The pre-specifiedsecondary outcomes were fatal or non-fatal cardiacarrest and all-cause mortality. The goal of REFINEwas to determine if noninvasive techniques couldbe used, alone and in combination, to define apopulation in whom an implantable defibrillatormay be of benefit.

Professor Cohen commented that REFINE “did notreport T-wave alternans (measured using eithermethod) as a univariate predictor of our secondaryoutcome of fatal or non-fatal cardiac arrest”.(Cohen R - 02/28/2008) This is not accurate. Asindicated in our paper, the results for thesecondary outcomes (fatal or non-fatal cardiacarrest and all-cause mortality) were similar tothose of the primary outcome. These data were notincluded in the manuscript do to page limits. Toassist in your decision making process these dataare provided in the following Table. The hazardratios in this Table reflect the independentassociation (risk ratio) of repolarizationalternans at the 3-month assessment, adjusted forejection fraction, age, gender, history ofdiabetes, and history of MI prior to the indexevent; the same statistical model used for theprimary analysis in REFINE. It is clear from thesedata that the spectral and MMA methods providesimilar information on risk, regardless of theoutcome evaluated.

Table. Secondary Outcomes in REFINE.

Fatal or Non-fatal Cardiac Arrest (24 events)

Method Risk p Positive Ratio value Sensitivity Accuracy

Spectral 3.4 0.03 83% 11% MMA 3.2 0.03 83% 11%

All-cause Mortality (30 events)

Spectral 2.8 0.03 80% 13%

MMA 2.5 0.04 75% 11%

Professor Cohen also commented that REFINEreported “rather low correlation (kappa = 0.17)between the results of the MMA and spectralanalytic methods”. (Cohen R - 02/28/2008) However,he failed to include the next sentence from ourpaper, that “high agreement was observed betweenthe techniques among patients who developed theprimary outcome (kappa = 0.49, p = 0.009) andthose who died (kappa = 0.41, p = 0.02) orsuffered a cardiac arrest (kappa = 0.48, p =0.02)”. (J Am Coll Cardiol 2007;50:2275- 84) Thus,this data too supports the notion that thespectral and MMA methods identify the samepatients who are at risk for serious outcomes.

Moving Average Cut-points

The adoption or lack of adoption of any riskassessment methods should be judged on scientificdata. I agree with comments that the cut-point fordefining an abnormal MMA results vary between thereported studies. I have made this same comment.(Exner D - 09/05/2007) Professor Cohen speculatesthat these variable MMA cut-points “suggests thepossibility that any association of MMA measuredT-wave alternans with endpoint events could bespurious resulting from the retrospectivefitting”. (Cohen R - 02/28/2008) I respectfullydisagree with his assertion.

The different cut-points simply reflect differentapplications of the MMA method. It is clear fromprior studies that optimal cut-points for the MMAmethod principally depend on two factors: 1) thesetting in which MMA is assessed (e.g., ambulatoryrecording, exercise test, atrial pacing, orpost-exercise) and 2) the number of beats used forthe MMA calculation (e.g., 64, 32, or 16).

In situations where noise is more prevalent (e.g.,exercise or ambulatory recording) or a shorterseries of beats is used the cut-point will behigher due to the inherent properties of thismethod. As discussed by Dr. Verrier, multiplestudies using exercise or ambulatory recordingshave consistently shown that an MMA cut-point of47 uV predicts a higher risk of adverse outcomes.(Verrier R -03/12/2008) When MMA is evaluatedduring periods where noise is controlled markedlylower cut-points have been identified aspredictive of serious outcomes. MMA in thesesettings has been shown to be equivalent tospectral assessment in two independent studies(Pacing Clin Electrophysiol 2007;30:352-8 and J AmColl Cardiol 2007;50:2275- 84). Differences in MMAcut-points in these studies simply reflect thenumber of beats over which MMA was calculated andwhether or not noise was subtracted for. Ongoingcollaborative research that we have been involvedwith clearly indicates that highly consistentinformation is obtained when the MMA method isused in the same setting and over the same numberof beats.

Finally, from a scientific viewpoint one shouldquestion whether a single cut-point is in fact theoptimal means to implement either the spectral orthe MMA method. Research has shown that patientrisk increases as the level of alternans rises.(Ann Noninvasive Electrocardiol 2005;10:110-20)

Automation

Criticism of the MMA method due to a lack ofautomation is specious. Multiple diagnostictechniques that are commonly used in cardiology(e.g., ejection fraction assessment, exerciseassessment, coronary angiography, and perfusionimaging to cite only a few examples) are notautomated. Yet these tools provide valuableinformation used in the care of patients with orat risk of cardiovascular disease.

Final thoughts

In follow-up my prior comments (Exner D -09/05/2007), three independent studies have nowfound that spectral repolarization alternans alonedoes not reliably predict which patients with alow ejection fraction benefit from implantabledefibrillator therapy (SCD-HeFT, MASTERS 1, andCARISMA). This is truly concerning and raisesconcern whether any repolarization alternansassessment method should be used to aid theclinician on deciding whether to provide or deny agiven patient an implantable defibrillator.

History has taught us a valuable lesson that weought not ignore. Despite abundant observationaldata supporting the predictive value of othernoninvasive tools (e.g., heart rate variability,signal-averaged ECG), these techniques failed topredict which patients benefited from animplantable defibrillator when assessed in aproperly designed, randomized controlled, trial (NEngl J Med 1997;337:1569-75 and 2004;351:2481-8).

To date, neither the spectral nor the MMA methodhas been proven useful in guiding prophylacticdefibrillator therapy. Thus, it would seem logicalto either condone use of either the spectralmethod or the MMA method to identify patients atrisk of serious outcomes, since the data clearlyindicate they predict risk to a similar extent, orto decide that given the absence of randomizeddata, that neither of these techniques berecommended to guide implantable defibrillatortherapy.

Sincerely,

Derek V Exner, MD, MPH, FRCPC

Less

Having spent many years working with the mathematical methods of T-wave alternans analysis, I’d like to make the following point: The MMA and Spectral Analytic Method are substantively different, and should not be confused as equivalent methodologies. A key difference between the two methodologies is the measurement of the statistical significance of the alternans. The spectral method has a built-in measure to assess the significance of the measured alternans. This makes the spectral measure more robust. The MMA method creates a measure of alternans, but does not have an intrinsic means of distinguishing alternans that is “random” (i.e. due to a noise fluctuation) from alternans generated by the electrical activity of the heart . Since the two methodologies are different, MMA must be validated by a prospective studies that demonstrates its predictive value in the target population. It cannot be validated by a number of informal studies and a “me too” argument.

Less

I share my comments not through a doctorate nor degree. I am an investor in mtwa technology and have studied every aspect of the science behind it, its potential market, and most importantly, "market acceptance". All have relevance in the discussion ,you have asked for, here in this forum, and my knowledge on this subject grants me a voice on behalf of the 400,000 victims of SCD this year..I maintain cms is the only Pure Patient Advocate, who''s decisions , while Not affecting practice decisions, ultimately guides the fragile exchange between science and market.. We cannot continue to ignore the relationship, which is ironically necessary within the system, and CMS'' role in this exchange has no better example,than in the reimbursement NCD for MTWA.

I have invested in this technology because of the pivatol role it can play in lowering the rate of sudden cardiac death.

We now know that ICD''s themselves have not lowered the rate of scd. Theodore Chow explained very well, this time, why MTWA is a better predictor of Mortality than an ICD. If we allow ICD firings to be the standard by which all risk statifiers are measured, you will never find a test more predictive, because we know from Master1 that "90% of all VTE''s were icd shocks, and 50% of those are for non-lethal arrhythmias. So I ask the question how can MTWA prove its prowess if the standard by which you judge itself, is flawed?.CMS'' decision making must always be evidenced based but as a Government agency, and the last Pure Patient Advocate it must also look within these trials, such as MASTER 1, and utilize common sense, with a keen awareness of the conflicts of interest..Whether through malice or Human nature and relationships, the system cannot continue to bear the weight of unnecessary implantations, nor can the rate of sudden cardiac death afford "missed necessary implantations".

Theodore Chow wrote, "The key point is that the MASTER study showed that in ICD-treated patients, MTWA does not predict those more or less likely to receive shocks."..Well what relevance is there in testing patients already implanted?.None, but to take away its predictive value within a trial by using a standard which is right 50% of the time.Chow wrote, "MTWA is associated with total mortality risk, which is a pertinent piece of clinical information in determination of ICD candidacy."....Chow wrote, "The ALPHA study, unlike MASTER, had no ICD implantation in the study population, and is thus more applicable to a situation where the test is used in non-device treated individuals."..I believe the only goal CMS has is to test, non-device treated individuals..Therefore MASTER 1 is useless to the discussion and Alpha and ABCD answer the question before CMS.Chow wrote, "These effects would serve to cloud the association of MTWA with true arrhythmic risk in ICD treated patients."..

Why would a study that "clouds" the association form the basis for any decision?Mtwa ,again and again, over 10 years and 5 continents, has proven to be an integral part of a cardiogical workup of a potential victim of scd.

Which is why an entity like GE would like to grab a share of that potential market..Whether one believes MMA is the equivalent of MTWA, it does show, by the very aggressive approach GE has taken, that GE believes T wave alternans is a pertinent tool. I was quite pleased to see the new comments in favor of adding MMA, as it only enhances the scientific justification of using t wave alternans to risk stratify.. However it is has also been shown in study after study, that MMA has no Defined endpoint nor automated interpretation. It would be a mistake to allow for a roaming interpration of results.There already exists an opening for interpration with MTWA, by not mandating MTWA prior to implantation.A roaming interpretation of results would only further complicate, a referral channel of patients already neglected by a system split between doing whats best for patients vs, whats best for there relationships..Each of the now 17 comments posted, including my own, carry their bias''.Which why it is important the 18th comment, yours(cms),allows MTWA to continue its progress in convincing the EP community, that MTWA can solve the connundrum they now face, which I believe was caused, in part, by a flawed risk assessment strategy.

Less

GE Healthcare appreciates the opportunity to provide public comment on the CMS proposed decision for MTWA involving the modified moving average (MMA) method. Our letter outlines a number of issues that we urge CMS to consider with respect to the proposed decision. These include the following:

1) Lack of discussion of two key publications by Exner and Schusterman in the CMS memorandum;
2) Clarification of MTWA MMA technology attributes; and
3) Responses to CMS reservations regarding the Finnish Study.

Overall, we urge CMS to provide parity in its evidence requirements and decision criteria when evaluating alternative methods of MTWA. The Proposed Decision Memo explicitly states that “The question before us could be addressed either by independent demonstration of a clinical benefit of the MMA method of measuring MTWA or by demonstration of substantial comparability of the MMA and spectral analysis (SA) methods of measuring MTWA. The evidence we reviewed focused on the former.” The evidence CMS chose not to review was the Exner et al. study (JACC 2007), which demonstrates equivalent efficacy in a direct head-to-head study. Demonstrating substantial comparability would be in line with the FDA’s policy for clearance of the MMA method for measuring MTWA.

Evidence Considered in the CMS Review: According to the decision memo, 32 articles were considered for review. From this group of articles, CMS excluded unpublished articles, oral presentations, editorials, review articles, studies that did not specifically address MTWA as a risk stratifier for patients eligible for implantable cardioverter defibrillator (ICD) treatment and those lacking sufficiently detailed information on study design or discussion of results.

We are concerned that the studies published by Exner et al. (JACC 2007) and Schusterman et al., (Circ 2006) were not discussed in the CMS proposed decision memorandum (PDM). Moreover, the basis for this exclusion is unclear.

While we are not absolutely sure why these studies were not used in the analysis of this National Coverage Decision (NCD), based on the criteria listed above we can only assume these studies were not included because they “did not specifically address MTWA as a risk stratifier for patients eligible for ICD treatment” as these studies do not appear to fit any of the other criteria. We would like to point out the exclusion criteria used in this analysis as well as the original NCD(CAG-00293N,)is inconsistent with the CMS final outcome. For example, CMS analyzed the meta-analysis done by Gehi. They point out that the search criteria for articles included in the analysis were those that “provided primary data on results of MTWA and of clinical outcomes including sudden cardiac death (SCD), cardiac death, ventricular arrhythmias, and/or ICD shock.” Also, studies by Ikeda, et al., were included with primary endpoints of “sudden cardiac death, as well as VTE.” In both cases ICD benefit was either not primary endpoints or (in the case of the meta-analysis) not necessarily a primary endpoint. This seems to indicate that studies used in the original determination were not linked to ICD therapy.

The current Nationally Indicated Use for MTWA states, “Microvolt T-wave Alternans diagnostic testing is covered for the evaluation of patients at risk for SCD, only when the spectral analysis method is used.” This indicated use does not mention any specific treatment. Furthermore, in the final decision memo for the original consideration CMS wrote, “…we do not believe that the current evidence is sufficient to require that physicians use the results of MTWA testing to select appropriate patients for ICD implantation.” We believe CMS has correctly determined that measuring T-wave alternans is beneficial in evaluating the risk of patients for SCD. However, they specifically did not link this test to any particular treatment modality. We believe these studies provide vital information on the ability of the MMA method to properly identify those at risk for SCD, which is the nationally indicated use. In this context, we encourage CMS to reconsider their decision to exclude these two important studies that not only show the utility of TWA as a risk prediction tool, but also provide direct comparison of the MMA and Spectral method. Based on these studies, we believe that both methods of testing are substantially equivalent. Exner illustrates this in his study. He also stated this in the first comment period when he wrote, “Both the spectral method and the moving average method were independently associated with a 3-fold increased risk of cardiac death or non-fatal cardiac arrest.”

Understanding of Technology
In the proposed decision memo (PDM), CMS wrote, “In the MMA method of measuring t-wave alternans, electrodes placed on a patient’s chest are used to obtain 24-hour ambulatory EKG recordings using a Holter monitor.” Also, the critique of the Nieminen study twice questions the reason the researchers chose to use “standard EKG instead of a Holter monitor.” We would like to clarify that the indications for use we have filed in the 510(k)s for this device do not restrict its use to a specific device (or modality). Currently, we provide MMA on a Stress and Holter product, both with 510(k) on file.

Finnish Study
The CMS proposed decision memorandum includes a lengthy critique of the Finnish study. CMS placed significant weight on this study in their analysis, noting several reservations about its content. Given the importance of this study to the CMS analysis, we urge CMS to consider the following issues:

1) CMS criticized the ethnicity of the study population. Interestingly, however, many of the studies CMS used to support current coverage of the spectral method were not performed in the United States. In fact most studies found in the Gehi analysis as well as others (Hohnloser et al 2003, Ikeda et al 2000 & 2004) were conducted in Germany and Japan. (See the Decision Memo for Microvolt T-wave Alternans (CAG-00293N) Appendix A)

2) CMS remarks that ejection fraction (EF) was measured in approximately 50% of the population. While we agree that low EF identifies a high-risk population, we would also point out that it is widely known that those with slightly persevered EF are also at risk. Furthermore, the national indication for use does not require any particular EF when performing the test. Nieminen et al, clearly state, “measurement of ejection fraction (EF) is not routine for patients referred for a clinical exercise test” which was the cohort being studied. They did, however, provide this information on those patients for which it was available.

3) CMS further indicates that the cut-points were determined after the fact. To prove the contrary, Nieminen et al wrote, “Several cut points were evaluated, including the value coinciding with the 75th percentile (46 µV), the cut point that proved to predict cardiac arrest or arrhythmic death in a prior study..”. This pre-specified cut-point of 46 was statistically significant for predicting sudden cardiac death with an OR 2.9 (CI 1.2-7.1) p=0.02.

4) CMS notes an apparent discrepancy in the number of subjects reported as opposed to the total number of subjects from the study. We note that with the study’s requirement for 4-year follow-up, all additional data points had not yet been reported for the study prior to publication of results.

5) In the proposed memorandum, CMS wrote, “We are also not told what percent of patients were not willing to participate and how their demographic characteristics compare to those of the 1037 who gave consent; indicating possible mishandling of indeterminate findings and skewing of results.” We are not aware of this information being reported in any of the previously analyzed studies for the spectral method. Furthermore it would be impossible to gather this demographic or any other information from those who weren’t willing to participate given the study design.

6) CMS also notes that Nieminen et al. did not report why 3.4% of the subjects had an unsuccessful stress test. In fact, the authors wrote, “A test was technically adequate if storing the haemodynamic data and continuous digital ECG signal was successful.” Technical failures referred to the lack of stored ECG and blood pressure. Furthermore, we would also point out that the 3% “indeterminate” concern is less significant when compared to a 20-40% indeterminate rate for the spectral method (Kaufman 2006).

7) Finally, CMS points out that this study was not a randomized control trial. With reference to appendix A in the final decision for the original NCD, this critique is true of all studies used by CMS to support coverage of the spectral method. This study was, however, a prospective cohort study. According to appendix B in the final decision memo, CMS lists this type of study as the third highest in regard to methodological rigor.

In conclusion, we would ask CMS to reconsider its exclusion of the Exner and Schusterman studies. We would also request CMS clarify the description of the technology. The MMA method was specifically designed to accommodate a variety of host platforms. Lastly, we ask CMS to re-analyze the Nieminen et al. study and use the same criteria used with their previous decision. We believe there is substantial clinical literature to conclude that the MMA method is equivalent to the spectral method. GE Healthcare also believes that accurate identification of patients at risk for SCD is extremely important for increasing efficiency in the U.S. healthcare system. To this end, we would ask CMS to expand their coverage and not limit it to a single technology.
Sincerely,

Robert M. Honigberg, MD
Chief Medical Officer
lobal Medical Affairs and Clinical Strategy
GE Healthcare

1. Kaufman, E.S., et al., "Indeterminate" microvolt T-wave alternans tests predict high risk of death or sustained ventricular arrhythmias in patients with left ventricular dysfunction. J Am Coll Cardiol, 2006. 48(7): p. 1399-404.

Less

This letter is to express my opinion regarding CMS’s proposed decision memorandum that declines to reimburse physicians when they use the Modified Moving Average (MMA) method for T-wave alternans analysis to assess their patients’ risk for sudden cardiac death. The major emphasis of my clinical work has been the problem of sudden cardiac death and the use of ICDs for prevention of life-threatening arrhythmias.

At the outset, I would like to express my concern that in the present evaluation, important information regarding the proven predictivity of the MMA test was not considered. Specifically excluded was the investigation by Dr. Exner and coworkers (JACC 2007), who performed a side-by-side comparison of the MMA and spectral analysis methods in evaluating arrhythmia risk in post-MI patients who may be candidates for device therapy. This study is sizeable and controls for the variables that are considered in deciding whether or not to implant an ICD. The odds ratios and level of prediction for MMA were similar or superior to the spectral method.

A unique and advantageous feature of the MMA method is that it can be employed during ambulatory ECG monitoring, not solely during exercise. This option is important for multiple reasons, including the fact that in 20-40% of cases, spectral method exercise tests are indeterminate because of patients’ inability to achieve the requisite target heart rate. Only 3% of tests are indeterminate with MMA analysis. AECG-based TWA testing can also be readily combined with AECG-based measures of autonomic function, such as heart rate turbulence. There is no doubt that the future will belong to noninvasive multiparameter arrhythmia risk stratification, and accurate methods that can be embedded within routine AECG monitoring offer distinct advantages to both patients and physicians due to ease of use.

I also wish to comment on the prospect that TWA may be entirely eliminated from reimbursement as a result of the yet-unpublished MASTER I trial. It has become eminently apparent that there are fundamental flaws in the design of this investigation, most particularly, use of ICD firing as a surrogate endpoint for arrhythmia. Because the types of arrhythmias that activate devices can be self-terminating, relying on device firing as an indication of a life-threatening arrhythmia is misguided.

It is also underappreciated that ICDs themselves have the potential for proarrhythmia, as my colleagues and I discussed recently (Germano et al Am J Cardiol 2006;97:1255–1261). This effect is likely to have disrupted the predictive capacity of TWA in the MASTER study. Such weak evidence should not be used as a basis for discounting more than a decade of prospective cohort studies demonstrating the predictive strength of TWA by both spectral and MMA methods. Reversion to ejection fraction as a sole basis for device implantation would continue the excessive rate of ICD implantation, with well-recognized risks to patients.

Finally, the accumulated evidence from sizeable prospective clinical studies support the significant prognostic capacity of TWA assessed by the MMA method. In my opinion, it would be inappropriate to exclude the MMA method selectively from reimbursement, as it is a validated and useful method for physicians who care for CMS enrollees with heightened risk for life-threatening arrhythmias.

Mark E. Josephson, M.D.
Herman Dana Professor of Medicine, Harvard Medical School
Director of Harvard-Thorndike Electrophysiology Institute and Arrhythmia Services
Beth Israel Deaconess Medical Center, Boston, MA

Less

March 14, 2008

Steve Phurrough, MD, MPA
Director, Coverage and Analysis Group
Centers for Medicare and Medicaid Services
Mail Stop C1-09-06
7500 Security Boulevard
Baltimore, Maryland 21244-1850

Dear Dr. Phurrough:

Thank you for the opportunity to comment on the Centers for Medicare and Medicaid Services' (CMS's) proposed national coverage decision (NCD), Microvolt T-wave Alternans (CAG-00293R). America's Health Insurance Plans (AHIP) is the national association representing nearly 1,300 health insurance plans providing coverage to more than 200 million Americans. We are pleased to submit these comments on behalf of our members.

AHIP concurs with the CMS proposed decision not to expand coverage of microvolt t-wave alternans (MTWA) testing using the modified moving average (MMA) method due to a lack of sufficient evidence proving efficacy in evaluating patients at risk for sudden cardiac death. We also support CMS's more rigorous approach to analyzing data that validates the benefit of MMA on its own, instead of only comparing effectiveness with the spectral method.

In addition, we look forward to the final results of the MASTER 1 trial which has initially shown that risk stratification with MTWA using the spectral method currently covered by Medicare was not helpful in predicting life-threatening events. If those results confirm that MTWA is ineffective at determining which patients are at increased risk of sudden cardiac death, AHIP would support a full reconsideration of the NCD on MTWA using the spectral method.

Thank you for the opportunity to comment on this important issue.

Less

Dear Dr. Phurrough,

Cambridge Heart appreciates the opportunity to comment on the proposed decision memorandum for Microvolt T Wave Alternans (MTWA) diagnostic testing. We support the proposal to continue coverage of MTWA for patients at risk of sudden cardiac death, only when the spectral analytic method is used.

Since the original request to consider inclusion of the modified moving average (MMA) method in August 2007, two publications on MMA-TWA have emerged which provide further evidence to support the preliminary CMS decision.

Exner et al (2007) - The REFINE Study

Exner at al evaluated TWA using the both the MMA and spectral methods in post-MI patients with EF < 0.50 [1]. Spectral MTWA testing was conducted with the standard exercise protocol; MMA-TWA was measured using Holter recordings.

Pre-specified dichotomy limits for spectral MTWA and other variables were chosen based on the literature [2]. However, the authors indicate that there was a 'lack of prior data for the Holter TWA method [1].' For this reason, receiver-operator curves were generated and an optimal cutpoint was retrospectively selected. In REFINE, the optimal cutpoint was found to be 5.0 uV. This is significantly lower than previously reported MMA-TWA cutpoints reported in the literature which range from 10.75-65.0 uV [3-5].

While the MMA-TWA cutpoint was retrospectively fitted to be predictive of the primary outcome of cardiac death or resuscitated cardiac arrest [1], the overall agreement between the spectral and MMA-TWA methods was very low (kappa = 0.17, p = 0.002) [6].

In the proposed decision memo, several limitations of the MMA-TWA literature were cited including the fact that 'cutpoints were determined after the fact rather than prospectively' as well as the lack of a standardized cutoff point [7]. The REFINE study, a retrospective analysis that introduces a new voltage threshold for MMA-TWA, further reinforces these limitations. It is, however, a significant addition to the literature which supports the predictive value of spectral MTWA in post-MI patients.

Harvey et al (2008)

In a recent pilot study of 24 patients, Harvey et al showed that MMA-TWA derived from Holter recordings was strongly correlated with noise [8]. The authors indicate that 'noise significantly influenced [MMA] TWA values and may be introduced during many stages of the processing procedure [8].'

As suggested by one CMS commenter, noise control is an important factor in the comparison of measurement techniques for T wave alternans. The new study by Harvey et al supports previously published work which suggests that the MMA method is not as robust in the presence of 'non-white' noise common in ECG signals [9].

In contrast, the spectral method employs a sophisticated and comprehensive algorithm that incorporates several layers of noise reduction including: multi-segment electrodes which allow one to take advantage of the inherent redundancy in ECG signals, validated compensation mechanisms for motion and respiratory artifact using local and trans-thoracic impedance signals, and a built-in means for establishing the statistical confidence or certainty of the alternans measurement [10]. Without this last component, one may incorrectly conclude that significant alternans is present when it is not. Since the available MMA method does not currently incorporate such an evaluation, this may explain the strong positive correlation between noise and MMA-TWA observed by Harvey et al [8].

We believe that these recent publications provide further evidence that the two methods for measuring MTWA are not equivalent. As such, the substantial body of literature on the spectral analysis method is not applicable to the MMA method.

The MASTER Trial

In the proposed decision memorandum, CMS requested public comment on the reported findings of the MASTER 1 trial, specifically with regard to whether CMS should continue to cover MTWA in general, regardless of the method used [7]. The MASTER study has not been published in manuscript form in a peer-reviewed journal; therefore comments below are limited to data presented at AHA 2007 [11].

While the study limitations are not outlined in the abstract, several factors which may have impacted the outcome have been noted [12]. The first involves the primary endpoint: life-threatening ventricular tachyarrhythmic events (LTVTE) defined as arrhythmic cardiac death or 'appropriate' ICD discharge. In this case, the composite endpoint was dominated by ICD discharges which comprised over 90% of the endpoint events. However, it has been noted that ICD firings are a poor surrogate endpoint for sudden cardiac death (SCD) because shocks, even those deemed appropriate, often treat rhythms that would not have resulted in death [13-15]. This endpoint, therefore, tends to overestimate arrhythmic mortality, leading investigators to conclude that 'appropriate ICD shocks cannot be equated with aborted sudden cardiac death, as has been done in the interpretation of various nonrandomized [ICD trials] [15].'

Second, it has been suggested that the relatively low event rate may have limited the study's statistical power, thereby making a positive finding unlikely [12]. Excluding patients with previous EPS or MTWA tests likely biased the population against a high risk group, thereby lowering the incidence of endpoint events. Lastly, approximately 20% of patients in the MASTER trial received a CRT-D device. Because biventricular pacing can itself alter arrhythmic risk, this may also have impacted the study results [16].

The findings of the MASTER study are in clear contrast to numerous published studies ofspectral MTWA. Two studies that most closely resemble the MASTER cohort are those published by Hohnloser [17] and Chow [18]. Hohnloser et al analyzed 129 MADIT II-type patients and found that MTWA was a significant predictor of sudden cardiac death [17]. At 24 months of follow-up, no sudden cardiac death or cardiac arrest was seen among patients who tested T-wave alternans negative, compared with an event rate of 15.6% among the remaining patients.

In a study of 768 patients with ischemic cardiomyopathy and EF ≤ 35%, Chow et al found that MTWA was a significant predictor of arrhythmic mortality (HR = 2.29) as well as all-cause mortality (HR = 2.24) [18]. Over 85% of all patients had a history of MI, making this cohort very similar to the one studied in MASTER, and consistent results were obtained when the analysis was limited to patients with an EF ≤ 30%. The authors cite the statistical rigor of the study and note that MTWA remained an independent predictor even after controlling for demographics, clinical comorbidities, ICD status, and medication treatment. In addition, they emphasize that the 'increased risk associated with a non-negative MTWA result is mediated primarily through higher rates of arrhythmic mortality [18].'

Upon reviewing the cumulative studies on MTWA, it appears that the choice of a less clinically definitive endpoint (ICD shocks) is the primary factor that differentiates MASTER from other studies that have shown spectral MTWA to be a significant predictor of life-threatening ventricular tachyarrhythmic events. Previous studies of MTWA like those mentioned above included populations in which most patients did not have an ICD; in turn, ICD firings did not predominate as clinical endpoints. This is illustrated, for example, in the recently published ALPHA trial in which the vast majority of patients did not have an ICD [19].

Spectral Analysis MTWA Studies Published Since the Original 2006 NCD

Since the original National Coverage Determination (NCD) established in March 2006, several articles and abstracts have been published on MTWA using the spectral method. Some of the most clinically significant include ALPHA [19], ABCD [20], Chow [18] and Bloomfield [21]. Collectively, these four studies demonstrate the prognostic value of MTWA in over 2300 patients with LV dysfunction and significantly extend the scope of clinical evidence which supports the benefit of risk stratification using spectral MTWA. In addition, the largest study of spectral MTWA to date (n=1041) was published in 2006 by Ikeda et al [22]. This study showed that MTWA is associated with arrhythmic events in patients with a history of myocardial infarction and preserved LV function (EF ≥ 40%).

Two new studies, including ABCD, suggest that an MTWA-directed approach to ICD implantation in patients with LV dysfunction would improve the therapeutic efficacy of implantable defibrillators [20, 23]. In addition, a recent article by Chan et al illustrates that such an approach would also be substantially more cost-effective [24]. Daccarrett et al recommend that MTWA be routinely performed in patients with LVEF ≤ 40% to better assess the chance of patient benefit from ICD implantation [25].

Lastly, some of the most ground-breaking research on MTWA in the last several years has been done in the basic science realm. Such work has revealed cellular and subcellular mechanisms which not only demonstrate but explain the link between MTWA and arrhythmogenesis [26-33].

Conclusion

In closing, Cambridge Heart supports the proposal to continue coverage of MTWA for patients at risk of sudden cardiac death, only when the spectral analytic method is used.

The extensive body of work on MTWA strongly supports the unique prognostic value of spectral MTWA and continues to grow as illustrated by the studies cited here.

Thank you for your consideration of this matter.

Sincerely,

Ali Haghighi-Mood, PhD
Chief Executive Officer
Cambridge Heart, Inc.

References

Less

The following comments are offered in response to your Proposed Decision Memo for Microvolt T-wave Alternans (CAG-00293R). I am a Professor of Medicine at Northwestern University Medical School and also Adjunct Professor in the Department of Electrical Engineering and Computer Science. I have practiced clinical cardiac electrophysiology since finishing my fellowship in that field in 1979 and am Board of Internal Medicine certified in that field as well as Cardiology and Internal Medicine. I have a 23 year research collaboration with the engineering school, focusing on signal processing applications in the analysis of cardiac arrhythmias and sudden cardiac death. As a clinician and researcher, I have followed the long series of attempts since the mid-1970s to identify patients at high risk for sudden cardiac death. This became increasingly important as automatic defibrillators were developed as effective preventive therapy, with the parallel problem of the financial cost of widespread use of this technology. Proposed risk predictors have included history, physical exam, ECG, Holter, ejection fraction, exercise testing, signal-averaged ECG and more recently, T-wave alternans. Each of these methods can contribute to risk prediction, but low specificity has often been a limitation. In regard to T-wave alternans, you have reviewed and approved for coverage the concept, specifically of the spectral method. I was surprised by your initial negative decision on the modified moving average (MMA) method. I offer the following specific comments:
1) In your list of objections to the Nieminen study, you express concern about the study population in a number of ways. I would suggest that the broad population of all those getting an exercise test is an advantage and strength of the study, not a weakness. Many randomized trial are so carefully defined and focused on patients with highly specified characteristics that the results may only be applicable to a small subset of patients seen in routine clinical practice (and yet they are often generalized, appropriately or not.) A group getting exercise testing for all reasons seems to me an excellent test group, which can be expected to include patients with a wide range of other risk factors of sudden cardiac death and thus is reflective of what one would see in practice. Your objection that there is no specific control group is, it seems to me, answered to some extent by the broad and relatively non-selected nature of this large study.
2) The ethnicity of the study population in the Nieminen study was cited as an issue. I’m sure you don’t propose to approve the method only for Americans of Scandinavian heritage. While it is obviously desirable in NIH sponsored randomized trials to insure good representation of African Americans, women, and various other groups, I don’t agree that there is reason to believe the Finnish population is a poor representation of Americans at risk for sudden death. Is there any reason to believe, for example, that the word “Hispanic” in your objection refers to a population of any scientifically valid cardiovascular specificity? Taken to extreme, this attitude would cause us to ignore any clinical study not done in our own country.
3) The comment about preferring a “populations at higher risk” the multivariate analysis seems to me to address this. Further, one could make the argument that the method is likely to be even better in a population where the pre-test likelihood of sudden cardiac death is higher. Again, showing that the method works in a broader population seems to me to be an advantage more than a disadvantage, since it would likely better reflect expected clinical practice (see comment #1 above.)
4) I did not understand your comment that the Nieminen study started in 2001 but was not published until 2006. Isn’t that the time required to be able to report adequate follow-up?
5) I don’t see how the failure of 3.4% of the subjects in the Nieminen study to complete the stress test calls into any question the results in those who did. This percentage of excluded patients is extremely low compared with virtually all randomized clinical trials.
6) I would call to your attention the REFINE study (J Am Coll Cardiol 2007;50:2275–84) which also seems to show predictive value of the MMA method in a multivariate model. I thought this study was especially interesting when interpreted in light of the contrast between MADIT II and DINAMIT results. That is, risk prediction was good later after an MI and did not perform well early.
7) Mean patient age in REFINE was the early 60s, closer to Medicare, which seems to address one of your other reservations.
8) The paper by Shusterman et al (Circulation. 2006;113:2880-2887)also seems to me to contribute some confidence to this method of predicting risk. This study demonstrated that the phenomenon measured by TWA using Holter and the MMA method was strongly temporally associated with major ventricular arrhythmias. This suggests that TWA (measured by MMA on Holter data) may be centrally involved in the pathophysiology of sudden cardiac death.

For all of the above reasons, among others, I see no scientifically valid reason to exlude this particular method of measuring T-wave alternans compared with spectral methods.

Respectfully submitted,

Steven Swiryn, MD, FACC, FHRS

Less

I am writing in response to the proposed decision memo by the Centers for Medicare and Medicaid Services regarding the use of the Modified Moving Average method for T-wave alternans to assist in risk stratification for life-threatening arrhythmias. We have extensive experience with this analytical approach and employed it in our prospective analysis of more than 1000 patients, making it one of the largest investigations of TWA. As the memo indicates, our study was published in a leading international journal, namely, European Heart Journal 2007;28:2332-2337. We are concerned by the extensive unwarranted criticisms and misrepresentations with respect to study design, the integral intended role of TWA from the outset of the FINCAVAS program, and lack of appreciation of the relevance of the study population to the public health problem of sudden cardiac death.

Study design and methodology (criticisms # 1,5,7,8,9,10,11):

As our investigation was published as an original research article, it underwent extensive peer-review by outside experts as well as by the editors. Acceptance by the journal implies that the article was methodologically sound with respect to design and analysis and was relevant to sudden death risk stratification. In fact, it was deemed to be of special significance as it was selected as an “editor’s pick” and unlimited access was provided, gratis, online.

Given the careful scrutiny, we were surprised by the barrage of negative comments that were not founded on epidemiologic principles or were factually incorrect. For example, the study was misclassified as “consecutive series” when it is a prospective cohort study; the pre-specified cutpoint of 46 microvolts (stated on page 2334 of our article) was denied; the extensive multivariate analysis to account for potential confounding influences by 12 known cardiovascular risk factors (page 2334, Table 3) was also denied; and our comment that the small, 3% attrition rate was related to technical aspects of the test (page 2333) and not to patient health and therefore does not introduce bias into the results was not appreciated. The statement that the study has no control group is also not accurate, since odds ratios of 2.9 to 7.4 are reported (Table 3 on page 2334) based on data from the survivors, enrolled in the study, who are the appropriate controls.

We were concerned that the study was regarded as an afterthought, when we had designated TWA as an important endpoint in FINCAVAS in our published protocol (Nieminen et al. BMC Cardiovasc Disord 2006;6:9). We acquired the GE Healthcare exercise system, which is specifically designed to run the MMA TWA software, for this study. Another unwarranted criticism is that we did not include all available patients in our published study. Evidently, that’s simply because an extensive followup period had not been completed by all of the enrolled patients mentioned in the protocol.

Relevance of the findings (criticisms # 2, 3, 6)

It was suggested that the Finnish population may not be particularly relevant to the U.S. population. In fact, Finns share the unfortunate high incidence of sudden cardiac death that is observed in the U.S. Our population is no less relevant than those studied with the spectral method (summarized by Gehi et al J Am Coll Cardiol 2005;46:75-82), which involved predominantly European populations except for a few of the larger studies that were performed in Japan. It is noteworthy that the Japanese population has among the lowest risk of sudden cardiac death.

Concern was raised that the great majority of our patients had preserved rather than depressed left ventricular ejection fraction, as measured in 55% of the population. As we stated on page 2335, “It is probable that those in whom EF was not measured had even better cardiovascular health, because there was no need for EF determination. The literature states that ejection fraction is an arrhythmia risk stratifier only when EF levels are below normal.” In fact, most Americans over age 65 who die suddenly have preserved ejection fraction, and this is our study population.

Our publication was criticized for additional methodologic issues, particularly for not conducting a side-by-side comparison with the spectral method. Because our intent was to evaluate patients in the context of a routine, clinically indicated exercise test, it is not feasible to use the spectral method, which requires a nonstandard exercise protocol. Moreover, in this context it is worth stating that the odds ratios and predictive power of MMA TWA are comparable if not superior to those reported in the Gehi meta-analysis of spectral TWA studies (J Am Coll Cardiol 2005;46:75-82), as referenced in our article.

Sincerely,

Tuomo Nieminen, M.D., Ph.D., M.Sc.

Less

Our research group has been working both with the spectral and MMA method in the analysis of TWA. A methodological problem of spectral method has been that many of the patients with heart disease (post-AMI) are not able to reach a target heart rate (>90-105 bpm)needed for analysis of TWA. So far, our two studies (Tapanainen et al, published in the Journal of Cardiovascular Electrophysiology, and Huikuri et al, CARISMA study presented in thge Late-Breaking Trial Sessions of HRS in Denver 2007)could not show that TWA analyzed by spectral method will provide prognostic information among the patients with a recent AMI. Another study, where I have participated (FINCAVACS)showed that TWA analyzed by a MMA method can predict sudden cardiac death in a large unselected patient population undergoing exercise testing (published in the European Heart Journal by Nieiminen et al.). This is a Finnish population, which should be quite similar to a US white population in terms of cardiovascular mortality and risk factors. An obvious advantage that we observed was that a specific target heart rate was not needed to analyze the TWA with a MMA method. Another obvious advantage was that MMA method does not require high-resolution electrodes, which are rather expensive and are not very user-friendly. We are currently analzying the predictive value of MMA method in the CARISMA database, which is a multicenter study evaluating the predictive value of several known risk stratification tests among the patients with a recent AMI and LVEF < 0.40. This analysis will further clarify the clinical impact of TWA analysis by MMA method. Finally, clinicians worldwide really need other risk stratification methods beyond measurement of LVEF in estimation of the need of implantable-cardioverter defibrillator therapy to make this therapy attractive and cost-effective.

Less

Below is a letter that was sent by GE Healthcare to CMS on 21 September 07. It is with regards to comments made by Professor Cohen on 16 Sep 07, the last day of the first comment period.

This letter is in response to the public comments regarding the National Coverage Analysis for Microvolt T-Wave Alternans (CAG-00293R). After reviewing the comments, we feel there are a number of comments the require clarification. We appreciate the opportunity to respond to these public comments. In our rebuttal, we will provide evidence from the peer-reviewed scientific and clinical literature.

Comment 1: Dr. Cohen, a co-inventor of the spectral method (SM) for T-wave alternans (TWA) analysis, began his public comment by stating, “Merely because another method has the word ‘alternans’ in its name does not in any way indicate that it is measuring the same phenomenon with the same accuracy and precision, nor does it indicate that the two methods are in any way equivalent.”

RESPONSE: This comment implies that GE Healthcare is relying merely on an unsubstantiated use of the term alternans in the title of the method as constituting the sole evidence supporting Modified Moving Average (MMA) as a technique for quantifying TWA. Rather, there is extensive peer-reviewed evidence from the scientific (22-24), engineering (9,18), and clinical literature (4,5-7,25-28) demonstrating MMA’s high level of accuracy. As a result, the MMA method was cleared by the FDA to measure TWA with the identical 1-microvolt resolution as SM.

Two groups of investigators concluded that the SM and MMA analytical techniques detect the same electrophysiological phenomenon of alternation in T-wave morphology that is associated with arrhythmic events (4,26). In an analysis of all patients with monomorphic, sustained ventricular tachycardia enrolled in the NIH-sponsored Electrophysiologic Study versus Electrocardiographic Monitoring (ESVEM) trial, Shusterman et al (26) reported parallel increases in TWA levels with both MMA and spectral analyses at 10 minutes prior to onset of the arrhythmia.

Two other groups (25,28) published clinical trials in which MMA analysis of TWA stratified risk for lethal ventricular arrhythmias or cardiac arrest based on a priori cutpoints. MMA’s utility is supported by Nieminen and colleagues (25), who recently reported their findings from a study of 1037 consecutive patients referred for routine clinical exercise testing using standard electrodes. All of the tests could be analyzed by the GE system, as there were no indeterminate results. TWA’s predictivity with MMA was excellent, with a highly significant odds ratio of 7.4 and negative predictive power (NPV) of 98.6%. These results are at least similar to those found with SM. Use of MMA for TWA analysis in routine ambulatory ECGs is supported by a study by Exner et al (5,7), who found in post-MI patients that MMA exhibits comparable predictivity to SM.

Comment 2: Dr. Cohen further underscores the fact that TWA analysis with SM necessitates specialized electrodes, automated interpretation algorithms, and hundreds of thousands of lines of software code. Patients’ heart rates must be elevated and maintained within a specific band for a specific time.

RESPONSE: The fact that SM approach requires physicians to depart from the routine clinical exercise protocol to satisfy SM’s stationarity requirements, its high indeterminate test rate of 20 to 40-% (14), and the use of costly specialized electrodes prompted GE Healthcare to search for an alternative approach for TWA analysis, which would permit evaluation in the normal flow of clinical testing, not only for exercise stress testing, but also for ambulatory ECG. The time-domain analysis technique of MMA was selected by GE and built upon technology already developed by GE for evaluating repolarization changes during exercise or ambulatory ECG monitoring.

Since the early introduction of computerized exercise testing (1) GE has employed advanced signal processing methods for removing artifacts and measuring dynamic ST/T wave changes 11,12,27). For the purposes of removing muscle artifact, GE has used a weighted average (moving average) to generate representative cardiac complexes (templates) that can be used to measure changes in repolarization without the contaminant effects of muscle artifact (17). This methodology has been well established and validated (2,17,20). GE was also the first company to employ the cubic spline technique for removing baseline sway or respiration artifact (19). This achievement was further enhanced for the evaluation of T-wave alternans (13). As with SM, these advanced signal processing methods require accurate QRS detection and time alignment (30). And, indeed, these techniques require hundreds of thousands of lines of code and extensive innovation and experience in the area.

Both SM and MMA are based on the assumption that the wave of repolarization is alternating on every other beat. SM measures the individual amplitudes of repolarization in each beat across a continuous set of 128 beats. Assuming that each of these individual measurements represents a peak or trough in a continuous sinusoidal waveform, a Fast Fourier transform can be applied to convert the sinusoidal waveform into the spectral domain of cycles per beat. At ½ cycle per beat, SM measures the TWA amplitude. In contrast, MMA simply bifurcates the processing for the odd and even beats. It generates a separate moving average template for the odd versus even beats. In contrast to SM, MMA does not require an uninterrupted set of beats in order to process the ECG signal. It can remain in synchronization with the alternans phenomenon, while skipping over excessively noisy beats or premature ventricular complexes. As a result, MMA is more forgiving of the demands of the clinical environment. MMA also forms templates, with which physicians can scrutinize T wave morphology for potential artifact and noise as well as to obtain insights into the pathologic basis for TWA on an individual patient basis.

Comment 3: It is suggested that SM detects an extremely low-level phenomenon of 1.9 microvolts, which can be identified only by Cambridge Heart’s proprietary SM protocols and equipment and not by the MMA approach. Thus, there is no equivalency between SM and MMA because MMA does not accurately measure TWA in this low range, where it is clinically meaningful.

RESPONSE: The claim that clinically significant TWA occurs only in the range of 1.9 microvolts is unsubstantiated. In fact, there is no “gold standard” means to verify this level of TWA independently from clinical ECGs. Low-voltage electromotive potentials generated at this level are detected by SM and by MMA analysis with the same level of precision and with a linear correlation of >0.9 (9). SM also requires an alternans ratio (“K”), a measure of signal-to-noise, to be greater than 3 (3). This requirement is often the limiting factor for a positive test, because as noise increases, a larger TWA must be detected in order to exceed the requirement of K >3. By definition, ECGs that contain more noise will require a larger TWA in order to be considered positive. Therefore, in an SM test classified as positive when a patient is at rest, TWA is typically of lower amplitude than that measured during stress. When clinical TWA values are reported during exercise testing with SM, they are actually in the range of 9.1±5.8 (15) or 10.8±10.0 microvolts (standard deviation, SD) (16). Given this variability, a substantial number of TWA values with SM are in the 15- to 20-microvolt range, as is evident in a representative patient test report (Fig. 1 in reference 16). Because the MMA method detects a peak not an average TWA level as with SM, its reported values are, conservatively, 3-fold larger than with SM (9). Klingenheben’s reported mean values (15,16) thus correspond to those found with MMA by Nieminen et al (25) in patients with events, i.e., 45 to 56 microvolts, also during exercise testing. Therefore, the TWA values reported with the MMA method in patients with heart disease reflect elevated levels of cardiac electrical instability rather than noise.

A sizeable proportion of SM tests are classified as “non-negative” or “abnormal” based on factors other than the 1.9-microvolt criterion. Between 20 to 40% of SM tests are indeterminate due to noise and patient factors. They are classified as “non-negative” if a patient cannot (a) achieve the target heart rate, (b) has excessive premature beats, or (c) unsustained TWA (14). It can be argued that SM tests classified as “non-negative” do not constitute TWA tests, since no TWA values are obtained. Rather, the results relate to patients’ exercise capacity or ambient ectopy.

The inability of SM to measure unsustained but clinically important bursts of TWA may derive from its long update window of 128 beats, which is equivalent to a 1/32 update factor. Use of a more dynamic update factor, as is permitted by MMA, which reports peak values as updated for each 15-second window, is inherently better suited to assess changing levels of TWA in response to physiologic stimuli. In fact, Nieminen et al (25) demonstrated that MMA analysis is capable of measuring TWA in all technically successful exercise tests. Finally, as Exner reported in a side-by-side study in post-MI patients, TWA was significantly predictive with both MMA and SM analyses (5-7).

Comment 4: MMA measurements are contaminated by large amount of noise.

RESPONSE: The GE system provides a QRS-aligned TWA template to rule out the possibility that noise and artifact have produced an artificially high TWA level. This template superimposes the ECG strips to permit visual inspection and verification of TWA magnitude in terms of separation of the alternating morphologies of the T wave. If MMA measurement were substantially disrupted by noise, it would not have the high odds ratio and NPV, which are equivalent to or better than SM, as documented by Nieminen et al (25) and Exner et al (5,7).

Comment 5: There is no standardized cutpoint for MMA, and cutpoints reported vary widely. Also, there is concern that cutpoints were set retrospectively.

RESPONSE: Unlike the SM method, which is employed almost exclusively during exercise and occasionally during pacing, the MMA technique has been implemented for use in routine exercise (25) and ambulatory ECG (AECG) recording (7,28) as well as in pacing (4). It is important to choose cutpoints based on the type of the test. For routine exercise, the study by Nieminen et al (25) tested two cutpoints. The first, namely 46 microvolts, was based on the upper-quartile cutpoint derived by Verrier et al (28), which was found to be highly significant and was not retrospectively fit. The second cutpoint, 65 microvolts, was derived by optimization from the Cox regression analysis. An investigator who was blinded to outcomes conducted automated TWA analysis.

Exner and coworkers measured TWA by MMA from AECG records during a post-exercise rest period (5,7). The cutpoint of 5 microvolts was derived after subtracting an estimate of noise. Thus, the measured cutpoint was lower than that of the exercise study by Nieminen et al (25) but >5 microvolts. As was demonstrated in the original clinical TWA study by Verrier and coworkers (Fig. 2 in reference 28), the mean level of TWA in patients at rest is considerably lower than that observed during daily activity.

The study by Cox and coworkers (4) was performed during atrial pacing while its 41 patients were at rest. The lower MMA-TWA levels reported are accounted for by this setting as well as by technical factors including the use of customized rather than GE Healthcare’s commercial software and by the use of the less sensitive update factor of 1/16 rather than the standard 1/8. Had Cox et al (4) employed the 1/8 update factor, their results are likely to have been even more favorable to MMA.

Therefore, as is the case for arterial blood pressure measurement, the setting (i.e., rest vs. exercise) of TWA testing should be taken into account for accurate interpretation of reported values. This issue has emerged because the MMA method is compatible for use in AECGs as well as exercise. Narayan (4,21) and others have acknowledged that MMA’s suitability for use with AECGs is a potentially significant advance in the field of TWA testing. Exner’s demonstration (7) of the predictive power of AECG-based TWA in combination with autonomic parameters supports this view.

These comments regarding TWA cutpoint also address criticisms raised by Dr. Costantini and Mr. Haghighi-Mood. In particular, Dr. Costantini has indicated his plan to employ the MMA method in a side-by-side comparison in a study to guide ICD implantation. It would be appropriate to set the cutpoint according to the test platform, namely, exercise or AECG monitoring during rest or across 24-hours, as during daily activity, the TWA value in post-MI patients with events was nearly double that of the resting level (28).

Comment 6: The population studied in Nieminen et al (25) is not relevant to the ICD population, and predictivity of MMA in this study cannot be extrapolated to the ICD population.

RESPONSE: The Nieminen study (25) was designed to demonstrate the utility of MMA in the challenging low-risk population and clinical environment of routine exercise testing. It enrolled 1037 consecutive patients who were referred for a routine exercise test for a variety of reasons. The investigators reported that MMA-TWA predicted SCD with a relative risk of 7.4 (2.8-19.4, p<0.001) and NPV of 98.6%, predictivity results that are at least comparable to those reported for SM (8). Medications were maintained, and no special electrodes were required. The importance of this study derives from the fact that although the multi-risk subgroup population has a low, ~2% rate of sudden cardiac death, yet more than 200,000 individuals in this group succumb to sudden cardiac death annually (10). The utility of MMA in populations currently under consideration for ICD implantation, namely post-MI patients with reduced LVEF, was examined directly by Exner and colleagues (7). In this side-by-side comparison, the MMA method yields comparable results to SM.

Comment 7: Because of sensitivity in the range of 35% as reported by Nieminen et al (25), the MMA method could not be relied upon to rule out ICD implantation.

RESPONSE: The more conservative, prospective cutpoint of 46 microvolts employed in the Nieminen study (25) yields a sensitivity of 50%, even in this low-risk population. Exner et al (5,7) demonstrated comparable sensitivity between MMA and SM analyses. As discussed above, the NPV was excellent with both methods (7,25).

In summary, GE Healthcare maintains that there is extensive peer-reviewed evidence in the basic science, engineering, and clinical literature indicating that MMA accurately measures TWA and that the signal-to-noise ratio is robust. This assertion is supported by the predictive results reported by Nieminen et al (25), which is among the largest TWA studies published, and by Exner et al (7). The latter investigation is especially germane to the NCD reconsideration, as it involved a side-by-side comparison of the MMA with the SM method in the high-risk post-MI patient population in which critical decisions are made about ICD implantation.

Finally, we maintain that the current SM approach of basing the TWA test interpretation solely on the 1.9-microvolt cutpoint without incorporating TWA values represents a missed opportunity to gauge more completely the patients’ levels of risk. TWA quantification is within the capacity of SM, as advocated by Klingenheben and coworkers (16; see also 29), who have investigated TWA testing extensively. Quantitative TWA assessment, as incorporated into the MMA method, in addition to identifying patients with elevated arrhythmia risk, also permits evaluation of the efficacy of less costly medical therapies and other interventions that may help to lessen the cost burden of ICD overuse. GE Healthcare hopes that they will be given the opportunity to advance the utility of TWA testing.

REFERENCES

1. Bhargava V, Watanabe K, Froelicher VF. Progress in computer analysis of the exercise electrocardiogram. Am J Cardiol 1981;47:1143-1151.

2. Bigi R, et al. Exercise versus recovery electrocardiography in predicting mortality in patients with uncomplicated myocardial infarction. Eur Heart J 2004;25:558-564.

3. Bloomfield DM, Hohnloser SH, Cohen RJ. Interpretation and classification of microvolt T wave alternans tests. J Cardiovasc Electrophysiol 2002;13(5):502-512.

4. Cox V, Patel M, Kim J, Liu T, Sivaraman G, Narayan SM. Predicting arrhythmia-free survival using spectral and modified-moving average analyses of T-wave alternans. PACE 2007;30:352-358.

5. Exner DV et al. Noninvasive assessment of sudden death risk after myocardial infarction—results of the REFINE study. Heart Rhythm 2007;4(5):S94.

6. Exner DV. CMS public comment for microvolt T-wave alternans (CAG-00293R), 9/5/07.

7. Exner DV, Kavanagh KM, Slawnych MP, Mitchell LB, Ramadan D, Aggarwal SG, Noullett C, Van Schaik A, Mitchell RT, Shibata MA, Gulamhussein S, McMeekin J, Tymchak W, Schnell G, Gillis AM, Sheldon RS, Fick GH, Duff HJ for the REFINE Investigators. Noninvasive Risk Assessment Early After a Myocardial Infarction. The Risk Estimation Following Infarction, Noninvasive Evaluation (REFINE) Study. J Am Coll Cardiol, in press.

8. Gehi AK, Stein RH, Metz LD, Gomes JA. Microvolt T-wave alternans for the risk stratification of ventricular tachyarrhythmic events: A meta-analysis. J Am Coll Cardiol 2005;46:75– 82.

9. Hostetler B, Xue J, Young B, Kaiser K, Findeis M, Gutterman D. Detect short run of TWA event with time-domain algorithm. Comput Cardiol 2005;32:483–486.

10. Huikuri HV, Castellanos A, Myerburg RJ. Sudden death due to cardiac arrhythmias. N Engl J Med 2001;345:1473–82.

11. Kaiser W, Findeis M. Artifact processing during exercise testing. J Electrocardiol 1999;32 Suppl:212-219.

12. Kaiser W. Findeis M. Novel signal processing methods for exercise ECG. International Society for Bioelectromagnetism 2000;2:60-65.

13. Kaiser W, Findeis M, Young B. Improving T-wave alternans measurement quality by reducing noise and artifact. Computers in Cardiology 2004;31:445-448.

14. Kaufman ES, Bloomfield DM, Steinman RC, Namerow PB, Costantini O, Cohen RJ, Bigger JT. “Indeterminate” microvolt T-wave alternans tests predict high risk of death or sustained ventricular arrhythmias in patients with left ventricular dysfunction. J Am Coll Cardiol 2006;48:1399–1404.

15. Klingenheben T, Gronefeld G, Li Y-G, Hohnloser SH. Effect of metoprolol and d,l-sotalol on microvolt-level T-wave alternans: results of a prospective, double-blind, randomized study. J Am Coll Cardiol 2001;38:2013–9.

16. Klingenheben T, Ptaszynski P, Hohnloser SH. Quantitative assessment of microvolt T-wave alternans in patients with congestive heart failure. J Cardiovasc Electrophysiol 2005;16(6):620-4.

17. Lauer MS, et al. Exercise-induced QT/R-R-interval hysteresis as a predictor of myocardial ischemia. J Electrocardiology 2006;39:315-323.

18. Martinez JP, Olmos S. Methodological principles of T wave alternans analysis: a unified framework. IEEE Trans Biomed Eng (2005 Apr) 52(4):599-613.

19. Meyer CR, Keiser HN. Electrocardiogram baseline noise estimation and removal using cubic splines and state-space computation techniques. Comput Biomed Res 1977;10:459-470.

20. Mickelson JK, Veterans Affairs ACME Investigators, et al. Is computer interpretation of the exercise electrocardiogram a reasonable surrogate for visual reading?. Clin Cardiol 1997;20:391-397.

21. Narayan SM. T-wave alternans and susceptibility to ventricular arrhythmias. J Am Coll Cardiol 2006;47:269-281.

22. Nearing BD, Verrier RL. Modified moving average method for T-wave alternans analysis with high accuracy to predict ventricular fibrillation. J Appl Physiol 2002a;92:541-549.

23. Nearing BD, Verrier RL. Progressive increases in complexity of T-wave oscillations herald ischemia-induced VF. Circ Res 2002b;91:727-732.

24. Nearing BD, Verrier RL. Tracking heightened cardiac electrical instability by computing interlead heterogeneity of T-wave morphology. J Appl Physiol 2003;95:2265-2272.

25. Nieminen T, Lehtimäki T, Viik J, Lehtinen R, Nikus K, Kööbi T, Niemelä K, Turjanmaa V, Kaiser W, Huhtala H, Verrier RL, Huikuri H, Kähönen M. T-wave alternans predicts mortality in a population undergoing a clinically indicated exercise test. Eur Heart J, published on line July 25, 2007.

26. Shusterman V, Goldberg A, London B. Upsurge in T-wave alternans and nonalternating repolarization instability precedes spontaneous initiation of ventricular tachyarrhythmias in humans. Circulation 2006;113:2880-2887.

27. Simoons ML et al. Quantitation of exercise electrocardiography. Circulation 1981;63:471-475.

28. Verrier RL, Nearing BD, LaRovere MT, Pinna GD, Mittleman MA, Bigger JT, Schwartz PJ for the ATRAMI Investigators. Ambulatory ECG-based tracking of T-wave alternans in post-myocardial infarction patients to assess risk of cardiac arrest or arrhythmic death. J Cardiovasc Electrophysiol 2003;14:705-711.

29. Verrier RL, Kwaku KF, Nearing BD, Josephson ME. T-wave alternans: Does size matter? [Invited editorial on Klingenheben et al 2005]. J Cardiovasc Electrophysiol 2005;16:225-228.

30. Xue J. Rowlandson GI. Effect of noise on frequency and time domain methods of T-wave alternans. Computers in Cardiology 2004:449-452.

Regards,

Ian Rowlandson
Chief Engineer
Diagnostic Cardiology, Clinical Systems
GE Healthcare

Less

It is with great interest that I read the CMS decision on MMA TWA. The broad exclusion by the reviewers of studies not involving ICD-eligible patients appears arbitrary and ignores the well recognized fact that the majority of sudden deaths occur in post-MI patients with EF > 35%. Such a decision is also inconsistent with the CMS coverage decision on spectral TWA, which states that “Microvolt T-wave Alternans diagnostic testing is covered for the evaluation of patients at risk for SCD…” and does not specify that such evaluations should only take place on ICD-eligible patients. While risk-stratification for the high-risk post-MI patient is obviously important, the failure to consider the importance of risk-stratification in the much larger number of patients with only mild to moderate degrees of left ventricular dysfunction will certainly impact public health as a whole. With this in mind, the review by CMS failed to recognize important data when deciding the coverage for MMA TWA. Though the bibliography lists the REFINE study by Exner et al, no mention is made of this paper in the coverage discussion. A review of this study would reveal a strikingly similar hazard ratio for cardiac death or resuscitated cardiac arrest for both spectral and Holter-based MMA TWA. The addition of MMA TWA to findings of abnormal heart rate turbulence and LVEF < 50% provided as-good-as or better predictive value than spectral TWA plus these other non-invasive measures. Similarly, the paper by Shusterman and colleagues, also referenced but not discussed, provides important evidence that TWA as measured by the MMA method significantly increases prior to ventricular tachyarrhythmias, thus providing an important link between MMA TWA and clinical endpoints. In my own experience using serial MMA TWA testing in hemodialysis patients, it is clear that this technology may have particular use in the assessment of patients unable to exercise or in the assessment of transient risk factors or interventions intended to decrease the risk of sudden death. While I fully appreciate the limited predictive value of any single risk stratification technique, I do not believe that the available data supports the decision of CMS to deny coverage of MMA TWA. I sincerely hope that CMS reviews all pertinent data pertaining to this issue prior to making a final decision.

Less

I am writing to comment on the proposed CMS decision memo regarding coverage for T-wave alternans (TWA) testing using the Modified Moving Average (MMA) method. As a co-inventor together with Dr. Bruce Nearing (1), I would like to address misconceptions about the clinical literature as well as the technology.

First, it has been stated that there are no prospective clinical studies supporting MMA-based TWA analysis for sudden death risk stratification. In fact, three prospective clinical studies (2-4) provide evidence that MMA predicts sudden cardiac death with odds ratios that are comparable or superior to those of the collected spectral method (SM) literature (5). It has also been claimed, incorrectly, that cutpoints were set retrospectively in each of these studies and that there are no accepted criteria for test interpretation. The relevant studies will be discussed in the context of the technology.

The following are a few of the key methodologic considerations that merit clarification with respect to both the SM and MMA methods.

1. TWA CUTPOINT FOR ARRHYTHMIA RISK STRATIFICATION

SM: The general perception is that the spectral method utilizes a hard and fast metric that has remained stable across all published studies, namely the cutpoint of >1.9 microvolts (μV). When this criterion is met, the test is considered “positive.” When it is not met, the outcome is classified as “negative.” However, in 20-40% of cases (6), a satisfactory TWA test cannot be performed, and the result is classified as “indeterminate.” This relatively high indeterminancy rate has been a persistent problem, and has led to the development of complex, multi-step interpretation and classification schemes (7). Not until two years ago was there a systematic investigation of “indeterminate” TWA testing with SM to propose a basis for classification of such tests as “non-negative” or “abnormal.” The study concluded that an “indeterminate” test should be considered “non-negative” in three conditions: (a) Patients’ failure to reach the target heart rate of 105-110 beats/min, which occurred in 51% of indeterminate tests; (b) Occurrence of short bouts of TWA of unspecified duration, which occurred in 10% of cases; (c) Presence of excess ectopy (6). Such “indeterminate,” “non-negative” tests are considered “positive” in terms of guiding therapy, since they predict mortality at least as well as do positive tests. Importantly, no TWA measurement is actually made in these “indeterminate” tests. In the 51% of cases attributable to failure to reach target heart rate, the characteristic that is measured is exercise capacity or chronotropic incompetence but not TWA.

MMA: In 2003, we published the first clinical study (2), a TWA substudy of the ATRAMI study, to determine the capacity of MMA in ambulatory ECG (AECG) records to predict lethal arrhythmias or cardiac arrest in post-MI patients with relatively preserved left ventricular ejection fraction. As the cutpoint was unknown, we prespecified that we would employ the 75th %ile of TWA in the controls (survivors), as guided by established epidemiologic principles. This cutpoint proved to be 46.6μV in lead V1 and 53.0μV in lead V5. As explained in the published study, the higher cutpoint in lead V5 was consistent with the facts that TWA is a regionally specific phenomenon. Using the 46.6-μV cutpoint, we found in blinded analysis that MMA TWA was highly predictive of lethal arrhythmias or cardiac arrest, with an odds ratio of 4.2 (95% CI = 1.1-16.3). TWA at 53.0 microvolts in lead V5 was also found to be predictive, with an odds ratio of 7.9 (CI = 1.9-33.1). This level of confidence intervals (CI) is consistent with the fact that this was a relatively small study but compares favorably with the SM studies discussed in the Gehi meta-analysis (5), which reviewed studies in a total of 2600 patients, with 16 patients enrolled in the smallest.

Subsequently, Nieminen and coworkers (3) in a full-cohort prospective analysis of the FINCAVAS database tested the ATRAMI-derived cutpoint of 46μV and also explored other values to establish whether during routine exercise a different cutpoint might be optimum. They found that the prespecified 46μV cutpoint was highly predictive, with a significant odds ratio of 2.9 and with a more narrow confidence interval of 1.2-7.1 in this sizeable, 1000+-patient study. The cutpoint of 65μV appeared to provide better prediction in terms of odds ratios, potentially because the analysis was performed during routine exercise testing. In the FINCAVAS study, MMA supplied TWA values in all but 3% of tests, compared to the reported 20-40% of indeterminate results with SM (6).

Most recently, Stein and coworkers (4) studied the predictive power of MMA-based TWA in AECG records in post-MI patients with left ventricular dysfunction enrolled in the EPHESUS study. These investigators employed the pre-specified cutpoint of 47μV (rounding 46.6 to 47) and again found that this cutpoint was highly predictive of sudden death, with a significant odd ratio of 5.2 and confidence intervals of 1.8 to 14.6. It is of interest that when ROC curves were used to identify an optimal cutpoint, this value turned out to be identical to the ATRAMI-derived cutpoint.

Thus, in three clinical TWA studies with MMA analysis (2-4), the prespecified 46-47-μV cutpoint consistently predicted risk for sudden cardiac death with odds ratios that are comparable or superior to that of the aggregate SM literature (5).

2. IMPORTANCE OF UPDATE FACTOR

There has been considerable confusion about the different TWA values obtained using the SM and MMA analysis. The key importance of the update factor for measuring TWA has been underappreciated not only for MMA but also for SM analysis. This factor defines the fraction that the incoming beat can contribute to the mean TWA value. Accordingly, an update factor of 1/8 is more sensitive than 1/32.

2A. STUDIES EMPLOYING THE 1/32 UPDATE FACTOR

Traditionally, SM has employed the functional equivalent of a 1/32-update factor, as it requires the accumulation of 128 beats or >1 min of ECG data at a heart rate of >105 beats/min (6) to meet the requirement of sustained TWA. As discussed above, this requirement led to a significant number of “indeterminate” tests and to a change in interpretation of TWA test results, namely, to consider that unsustained episodes of TWA are prognostically significant (6).

SM’s standard use of the 1/32-update factor probably accounts in large measure for the low TWA values reported as compared to the MMA method. The sizeable influence of update factor on determined TWA values can be as great as 3-fold. In the FINCAVAS study, when 1/32 was employed, the average TWA values were in the range of 20-30μV compared to 50-60μV with 1/8 update factor. The latter proved to be superior in the prediction of sudden death. This low average value of 20μV for MMA with the 1/32 update factor in the FINCAVAS study are comparable to those reported in one of the few studies in which TWA was quantified with SM. Klingenheben et al (8) reported a mean TWA value of 10.8±10.0μV (SD), as discussed in our accompanying editorial (9).

With SM, other factors also decrease the reported TWA magnitude, as this approach averages not only (a) across >1 min but also (b) across the JT interval and (c) with respect to a mean beat rather than by comparing odd and even beats to each other. The MMA method by contrast seeks the maximum TWA level within the JT interval and updates the calculation every 15 sec (1). These considerations account for the fact that in the few SM studies with reported TWA values (8), the mean result is in the range of 10-20μV by comparison with 40-60μV by the MMA method.

2B. STUDIES EMPLOYING THE 1/16 UPDATE FACTOR

The prospective REFINE study by Exner and colleagues (10) utilized an update factor of 1/16 rather than 1/8, subtracted computer-generated values of noise, and used a considerably lower optimized cutpoint of 5μV. They found, nevertheless, in a side-by-side comparison that MMA TWA stratifies risk for cardiovascular death or resuscitated cardiac arrest in post-MI patients with depressed left ventricular dysfunction equivalently to SM. The odds ratios for independent prediction by TWA were similar for SM [2.75 (CI: 1.08-7.02, p=0.034] and MMA [2.94 (CI: 1.10-7.87, p=0.031]. Twenty-four of 29 patients with these endpoints experienced fatal arrhythmias or resuscitated cardiac arrest.

Cox and coworkers (11) used customized software and employed a 1/16 update factor in studies involved pacing rather than exercise or AECG monitoring. These considerations are likely to account for their lower optimized cutpoint of 10.75μV. It is quite possible, although unproven, that using higher-resolution GE Healthcare commercial MMA algorithms with the 1/8-update factor would improve predictivity, which was nearly significant (p<0.06).

A physiologic factor that may have contributed to a lower cutpoint level in both studies is that patients were supine and not undergoing exercise testing.

2C. STUDIES EMPLOYING THE 1/8 UPDATE FACTOR

As discussed above, when the original and recommended 1/8 update factor was employed, as in the post-MI AECG studies of ATRAMI (2) and EPHESUS (4) and in the exercise-based FINCAVAS (3) study in a low-risk patient population, the MMA literature has consistently shown that a TWA cutpoint of 46-47 μV stratifies risk for sudden death with odds ratios that are comparable to SM analysis (5). This sensitive update factor may contribute to MMA’s suitability to detect episodes of unsustained TWA, which have only recently been appreciated as prognostically important in SM studies (6).

SUMMARY

The firm scientific foundation of the MMA method, published in leading journals, and the sizeable clinical database of patients studied with this technology support the utility of MMA TWA for arrhythmia risk stratification. Odds ratios in clinical studies of MMA are similar to superior to those with SM (5). Furthermore, MMA’s compatibility with routine exercise testing and AECG recordings allows it to be used in the typical flow of patient evaluation. The method’s suitability for analyzing TWA in AECG records also allows testing in many patients who cannot exercise. Such CMS patients are probably the sickest and most likely to benefit from accurate TWA testing but cannot be tested by SM analysis.

This collective body of evidence, demonstrating MMA’s equivalence or superiority to SM, appears to warrant reconsideration by CMS of their proposed decision to exclude MMA testing selectively from coverage.

Sincerely,

Richard L. Verrier, Ph.D., F.A.C.C.
Associate Professor of Medicine, Harvard Medical School
Department of Medicine, Beth Israel Deaconess Medical Center, Boston

REFERENCES
1. Nearing BD, Verrier RL. Modified moving average method for T-wave alternans analysis with high accuracy to predict ventricular fibrillation. J Appl Physiol 2002; 92:541-549.
2. Verrier RL, Nearing BD, LaRovere MT, Pinna GD, Mittleman MA, Bigger JT, Schwartz PJ for the ATRAMI Investigators. Ambulatory ECG-based tracking of T-wave alternans in post-myocardial infarction patients to assess risk of cardiac arrest or arrhythmic death. J Cardiovasc Electrophysiol 2003; 14:705-711.
3. Nieminen T, Lehtimäki T, Viik J, Lehtinen R, Nikus K, Kööbi T, Niemelä K, Turjanmaa V, Kaiser W, Huhtala H, Verrier RL, Huikuri H, Kähönen M. T-wave alternans predicts mortality in a population undergoing a clinically indicated exercise test. Eur Heart J 2007; 28:2332-2337.
4. Stein P, Sanghavi D, Domitrovich P, Deedwania P. Ambulatory ECG-based T-wave alternans predicts sudden cardiac death in high-risk post-MI patients with left ventricular dysfunction in the EPHESUS study. J Cardiovascular Electrophysiol, in press; abstract, J Am Coll Cardiol 2008.
5. Gehi AK, Stein RH, Metz LD, Gomes JA. Microvolt T-wave alternans for the risk stratification of ventricular tachyarrhythmic events. J Am Coll Cardiol 2005; 46:75-82.
6. Kaufman ES, Bloomfield DM, Steinman RC, Namerow PB, Costantini O, Cohen RJ, Bigger JT. “Indeterminate” microvolt T-wave alternans tests predict high risk of death or sustained ventricular arrhythmias in patients with left ventricular dysfunction. J Am Coll Cardiol 2006; 48:1399–1404.
7. Bloomfield DM, Hohnloser SH, Cohen RJ. Interpretation and classification of microvolt T wave alternans tests. J Cardiovasc Electrophysiol 2002; 13:502-512.
8. Klingenheben T, Ptaszynski P, and Hohnloser SH. Quantitative assessment of microvolt T-wave alternans in patients with congestive heart failure. J Cardiovasc Electrophysiol 2005; 16:614-618.
9. Verrier RL, Kwaku KF, Nearing BD, Josephson ME. T-wave alternans: Does size matter? [invited editorial]. J Cardiovasc Electrophysiol 2005; 16:225-228.
10. Exner DV, Kavanagh KM, Slawnych MP, Mitchell LB, Ramadan D, Aggarwal SG, Noullett C, Van Schaik A, Mitchell RT, Shibata MA, Gulamhussein S, McMeekin J, Tymchak W, Schnell G, Gillis AM, Sheldon RS, Fick GH, Duff HJ for the REFINE Investigators: Noninvasive risk assessment early after a myocardial infarction. The Risk Estimation Following Infarction, Noninvasive Evaluation (REFINE) Study. J Am Coll Cardiol 2007; 50:2275-2284.
11. Cox, V, Patel M, Kim J, Liu T, Sivaraman G, Narayan SM. Predicting arrhythmia-free survival using spectral and modified-moving average analyses of T-wave alternans. Pacing Clin Electrophysiol 2007; 30:352-358.

Less

I am a Cardiac Electrophysiologist in a privatepractice setting(single specialty group)and wouldlike to comment on CAR-00293R. I am the MedicalDirector of a Clinical Research center and haveused MTWA for the past 8 years both in clinicaltrial work as well as routine practice.

I would like to support the CMS decision not torevise the MTWA NCD.

The initial decision to provide MTWA coverage wasbased on review of over 10 years of scientificstudies on MTWA by the Spectral AnalyticMethod.The development of MTWA analysis by the MMAmethod is promising (REFINE trial) but has notundergone rigorous scientific scrutiny as theSpectral Analytic Method has been. The potentialto cover a ew method that has not been vetted byclinical science would undermine the field ofnon-invasive risk stratification.

Last, CMS has requested comment on the MASTERtrial as it pertains to MTWA coverage.

The MASTER trial has only been published inabstract form and has not undergone true peerreview to assess the validity of this work.

I have a number of serious concerns on this studyand believe it should NOT impact the current MTWA NCD.

First, MASTER trial was conducted in very illpatients (MADIT II) with an average EF of 24% thusa group not normally undergoing MTWA before ICDimplant. MTWA did predict mortality but did notpredict ICD discharge. The use of ICD discharge asa surrogate marker for SCD is a serious flaw sinceat least 30% of ICD discharges are related tonon-fatal arrhythmia. I believe on that fact alonethis trial should be dismissed.

Secondly, MASTER trial was funded and coordinatedby INDUSTRY(Medtronic)with serious corporateinterests and conflicts with the outcome. The fearof reduction in the ICD implant market has drivencorporate interests to downplay non-invasive riskstratification despite the well-know limitationsof EF determination. A major corporate competitorhas now aligned with the company that hasdeveloped MTWA and this certainly has changed thelandscape

I would hope CMS continues to allow science toinfluence the NCD process as it has before Thanks

Less

To whom it may concern,

I am writing this letter in response to the CMS request for public opinion and comment on the memo with respect to coverage for microvolt T-wave alternans testing (MTWA) using the spectral method. In that memo CMS asked for comment on more recent data, specifically citing the MASTER Trial. As you know, the MASTER Trial was a prospective multi-center study of MADIT-II indicated patients, all of whom underwent ICD implantation following MTWA testing. This study showed no difference in the “ventricular tachyarrhythmic events” (VTEs) according to MTWA classification [1]. For those who have been skeptical about MTWA (and/or risk stratification in general), the MASTER Trial results have served to confirm their opposition.

As the principal investigator of MASTER, I stand by the quality and findings of this study. However, the real meaning of the MASTER study is contained in its details, which unfortunately are not broadly understood. In this regard, I have a few comments:

1) All patients were ICD-treated, and the endpoint was a composite one which included ICD shocks. In fact, 90% of all VTEs were ICD shocks – very little arrhythmic mortality was seen. We know from studies like the Ellenbogen review of the DEFINITE study that 50% of ICD shocks are for non-lethal arrhythmias [2]. With this background of ICD shocks, some meaningful and some not, it can be difficult to extrapolate anticipated mortality differences in non-ICD treated patients. The key point is that the MASTER study showed that in ICD-treated patients, MTWA does not predict those more or less likely to receive shocks. Extrapolation of the interpretation beyond this point – including to non-ICD treated patients – should only be done with great caution.
2) The MASTER study is consistent with many other studies in showing that MTWA is associated with total mortality risk, which is a pertinent piece of clinical information in determination of ICD candidacy. The ALPHA and ABCD studies, the results of which were released after the initial CMS decision, confirm MTWA’s predictive ability in both ischemic and non-ischemic prophylactic ICD populations [3, 4]. The ALPHA study, unlike MASTER, had no ICD implantation in the study population, and is thus more applicable to a situation where the test is used in non-device treated individuals [4]. This study showed excellent survival without a defibrillator in non-ischemic cardiomyopathy patients.
3) More interestingly, Ikeda and colleagues have recently shown in post-infarction survivors with relatively “preserved” EF, that MTWA predicts outcome with a hazard ratio of nearly 20 [5]. Application of MTWA within this population is intriguing. It may be, as Rashba and colleagues have suggested, that MTWA actually performs better in patients with lesser degrees of LV dysfunction [6]. Therefore, CMS support of coverage in this (higher EF) population is also justified.
4) Clinical studies, including those from our group, have shown that non-ICD treated MTWA negative patients have a mortality risk similar to ICD-treated patients [7, 8]. These studies eliminate the “over-treatment” effect from the ICD. A propensity score analysis of ICD and non-ICD treated MTWA negative patients that we published previously in JACC confirmed the above point [7].
5) One needs to consider that some arrhythmias are actually induced by the presence of the ICD, either through pacing (especially long-short facilitated sequences), or possibly tip extra-systoles [9]. The former has been studied and quantified by Sweeney and colleagues [10]. These effects would serve to cloud the association of MTWA with true arrhythmic risk in ICD treated patients. This needs to be considered in the evaluation of any study using ICD shocks as an endpoint.

The MASTER Trial has been very useful in helping us understand how to frame questions with respect to risk stratification in ICD-treated patients. It has taught us what MTWA can and cannot do in this population. I fully stand by our findings. The question really is, though, what pertinence this has to CMS in a decision about MTWA testing in a population of patients without ICDs.

To summarize, then, I believe that CMS should stand by its original decision with respect to MTWA coverage of MTWA using the spectral analytic method, and that there is not sufficient data – inclusive of the MASTER Trial – to recommend reversing this coverage policy.

Thank you for your attention.

Regards,

Theodore Chow MD, FACC, FAAC

Ohio Heart & Vascular Center
Lindner Clinical Trials Center and
Principal Investigator, MASTER Trial

1. Chow, T., et al., Primary Results From the Microvolt T Wave Alternans Testing for Risk Stratification of Post MI Patients (MASTER I) Trial. Circulation, 2007. 116(22): p. 2631.
2. Ellenbogen, K.A., et al., Are implantable cardioverter defibrillator shocks a surrogate for sudden cardiac death in patients with nonischemic cardiomyopathy? Circulation, 2006. 113(6): p. 776-82.
3. Constantini, O. and D.S. Rosenbaum, ABCD (Alternans Before Cardioverter Defibrillator). Clin Cardiol, 2007. 30: p. 97-100.
4. Salerno-Uriarte, J.A., 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(19): p. 1896-1904.
5. Ikeda, T., et al., Predictive value of microvolt T-wave alternans for sudden cardiac death in patients with preserved cardiac function after acute myocardial infarction: results of a collaborative cohort study. J Am Coll Cardiol, 2006. 48(11): p. 2268-74.
6. Rashba, E.J., 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(2): p. 170-6.
7. Chow, T., et al., Microvolt T-wave alternans identifies patients with ischemic cardiomyopathy who benefit from implantable cardioverter-defibrillator therapy. J Am Coll Cardiol, 2007. 49(1): p. 50-8.
8. Chow, T., et al., Prognostic utility of microvolt T-wave alternans in risk stratification of patients with ischemic cardiomyopathy. J Am Coll Cardiol, 2006. 47(9): p. 1820-7.
9. Germano, J.J., et al., Frequency and causes of implantable cardioverter-defibrillator therapies: is device therapy proarrhythmic? Am J Cardiol, 2006. 97(8): p. 1255-61.
10. Sweeney, M.O., et al., Bradycardia Pacing-Induced Short-Long-Short Sequences at the Onset of Ventricular Tachyarrhythmias: A Possible Mechanism of Proarrhythmia? J Am Coll Cardiol, 2007: p. j.jacc.2007.02.077.

Less

I am a researcher in signal processing applied to ECG signal. Having worked in TWA detection and estimation, I would like to contribute to this discussion, highlighting some aspects of the proposed decision memo PDM from a signal processing point of view.

First, I found some misconception of the term "stationarity" on the PDM.

For instance, in the sentences "The spectral analytical method requires the patient to be stationary when obtaining the data, thus, this data cannot be acquired with ambulatory ECG monitoring such as Holter monitors. In the MMA method of measuring t-wave alternans, electrodes placed on a patient’s chest are used to obtain 24-hour ambulatory EKG recordings using a Holter monitor."

Stationarity is a property of stocastically-modelled signals meaning that their statistical properties do not change with time. While this definition is not accomplished by any physiological signals, the word "stationary" refers usually to "short-term stationarity", i.e. that the signal properties are stable within a given time period. Thus, stating that the "patient is required to be stationary" is not correct. In fact, as we discussed in (Martínez Olmos 2005, cited in the PDM), both the spectral and MMA methods require signal stationarity in a given time period wich is defined by some of their parameters.

In the same way both signal processing methods could be used to measure TWA in a stress test or in a Holter recording. Both methods should be seen as two ways to get the same goal: estimate the TWA magnitude in an ECG signal. Considerations about whether the spectral or MMA methods require that patient reach a specific heart rate is not a consideration to be done about the signal processing method, but to the phenomenon called TWA. It is known that TWA (as an ECG phenomenon, regardless of how it is quantified) may be unspecific at higher heart rates and unsensitive at lower rates, and this is the motivation for the HR requirement.

In the PDM, page 6, public comment is required about advantages and disadvantages of measuring TWA in stress test EKG or Holter monitor. From my experience, testing TWA in stress test has to deal with more noisy signals, and some sensitivity is lost because of it. However, heart rate may be controlled by the cardiologist. In the case of Holter recordings, signal is usually cleaner but heart-rate is spontaneous, which may also cause a loss in sensitivity. But both methods can be used to analyze stress and ambulatory signals.

In Martínez, Olmos (2005), signal processing considerations are made about both methods. When analyzing the same type of signals, I see no reason to favor one method with respect to the other. I think it would be more sensible to define guidelines about how and in which conditions the ECG signal should be acquired for a clinically useful TWA testing.

Less

There seems to be some confusion in the comparison of the spectral and time-domain methods for the analysis of T-wave alternans. In particular, it is well known (Parseval''s theorem) that the spectral and time-domain representations (of a square or a power of a Riemann-integrable function) are identical. Thus, the spectral and time-domain methods, in theory, should provide identical results. However, in a real-life setting, the results are not identical, due to the presence of multiple sources of noise and artifacts. Nevertheless, the results of the two methods are usually highly correlated (i.e. similar). Several clinical studies have compared those two methods and have shown similarity between the outputs of the two approaches.

Therefore, I think, the comparison of methods for analysis of T-wave alternans should focus on noise control (i.e. which method is more robust in the presence of different types of noises and artifacts, including baseline wander, respiration, movement artifacts, etc.). Both the spectral and time-domain approaches rely on a combination of noise-control procedures, which are not trivial. Thus, in my view, the most important question is: how much noise control can be achieved with a particular filtering or noise detection scheme and whether such filtering is sufficient for reliable quantification of T-wave alternans. There were a number of studies on this subject over the past 5-10 years that address this issue and should be taken into account.

Less

I would like to register my opinion regarding theproposed CMS decision not to cover the ModifiedMoving Average (MMA) method for TWA detection andmeasurement while continuing to cover theCambridge Heart Exercise-based spectral one. AsDirector of the Washington University School ofMedicine HRV Laboratory, I have been working inthe field of noninvasive arrhythmia riskstratification from ambulatory ECGs (AECGs) foralmost two decades. I have closely followed thedevelopment of the MMA method from its inception,from early mechanistic studies to more recentsizeable clinical trials by Nieminen et al (2007)and Exner and coworkers (2007). Because myprimary interest is in obtaining the maximumclinical information from AECGs, I was extremelyexcited when I found out that the MMA could beapplied to them. In our laboratory, we recentlycompleted an investigation using MMA TWA in AECGrecords of post-MI patients with left ventriculardysfunction from the EPHESUS study. This studywill be presented at the upcoming American Collegeof Cardiology meeting and the full-lengthmanuscript has been accepted for publication inthe Journal of Cardiovascular Electrophysiology.

Our findings demonstrate that using thepre-specified cutpoint of 47 microvolts, which isnearly identical to that reported by Dr. Verrierand colleagues for the TWA substudy of the ATRAMIstudy, MMA-based TWA from AECGs provides a highlysignificant odds ratio of 5.2 or greater forsudden death (SCD), with narrow confidenceintervals, supporting the capacity of this test todiscriminate and identify patients at risk forsudden death. Thus, our observations confirmthose of Verrier and coworkers (2003), discussedin the proposed decision memo, in terms of theutility of MMA and now constitute the third studyin which TWA at a cutpoint of 46 to 47μV predictslife-threatening arrhythmias with high odds ratios.

In my opinion, there is considerable evidence inthe literature, based on published reports and ourin-press study, which demonstrate that MMA-basedTWA is equivalent or, because it is permits adirect view of the exact T-wave changes ratherthan a black box approach, potentially superior tothe spectral method in assessing risk for SCD. Asour study was performed from AECG records, itdemonstrates how readily TWA can be assessedduring the normal flow of clinical care. Thesestudies were conducted in patient cohortscomparable to the CMS population, who can benefitfrom this analytical test. Accordingly, I urgeCMS to reconsider their decision and supportreimbursement for MMA-based TWA analysis.

Less

I am disappointed that CMS is not extending coverage of microvolt T-wave alternans (MTWA) to include the modified moving average (MMA)method. I am surprised that CMS did not use the REFINE study as part of their coverage determination (Exner et al. Noninvasive Risk Assessment Early after Myocardial Infarction. Journal of the American College of Cardiology 2007;50:2276). This study assessed MTWA in 322 high risk patients with reduced ejection fraction following myocardial infarction who were followed for a median of 47 months. MTWA along with impaired heart rate turbulence and reduced ejection fraction identified patients at risk for serious events. The study compared MTWA determined by exercise testing using the CMS approved spectral technique and MTWA assessed by Holter using theMMA technique. The hazard ratio for the two parameters to predict development of cardiac death or resuscitated cardiac arrest was actually higher for the MMA technique compared to the spectral technique at 10-14 weeks after the index myocardial infarction (HR of 2.94 for the MMA technique and 2.75 for the spectral technique). The optimal combination of parameters to predict an adverse cardiac event included ejection fraction under 50%, TWA using the MMA technique, and HRT measured at 10-14 weeks after myocardial infarction. The REFINE study was well designed and published in a highly respected cardiology journal. This data should have been used in the coverage decision assessment of the MMA method for determining MTWA. The use of the ICD in patients at high risk for sudden cardiacdeath has been demonstrated to prolong life. However, many patients with the ICD in place never use the device. Identifying which patients will receive the most benefit from the ICD has the potential to improve the usefulness of the device and decrease costs associated with this technology. Screening tests such as MTWA need to be available so that they can be studied and assessed in large populations of patients to determine if they are effective in identifying patients at both high and low risk for developing sudden cardiac death, and thus who may benefit from the ICD. The MMA method of assessing MTWA has been demonstrated to be equivalent to the spectral method for measuring MTWA and should be approved by CMS for coverage as both technologies are substantially equivalent.

Stephen C. Hammill, M.D.
Professor of Medicine Director, Electrocardiography Laboratory
Mayo Clinic

Less

My area of expertise is assessment of cardiovascular risk in patients with ischemic heart disease. Ambulatory ECG and exercise stress testing are essential noninvasive tools for evaluating major characteristics of risk, including both functional ischemia and susceptibility to life-threatening arrhythmias.

Over recent years, I have been particularly impressed by the growing evidence suggesting that tracking repolarization abnormalities with T-wave alternans can add significantly to risk profiling. Whereas the most commonly employed TWA analysis technique has been the spectral method, the more flexible Modified Moving Average method represents a significant forward step. The study by Nieminen et al (European Heart Journal, 2007), which is sizeable, enrolling more than 1000 patients, and reporting with highly significant odds ratios of 4 to 7, supports MMA’s utility in sudden death risk assessment.

It is highly significant that the MMA method can be used to measure TWA in ambulatory ECGs. Its suitability to predict arrhythmia risk was initially shown by Dr. Verrier and colleagues in the TWA substudy of the ATRAMI study (J Cardiovasc Electrophysiol 2003). MMA TWA’s predictivity was recently confirmed in studies in patients with depressed ejection fraction by Exner and coworkers (J Am Coll Cardiol 2007) and by Stein and coworkers, to be presented at the American College of Cardiology annual meeting. It is important to emphasize, as Dr. Exner has shown, that TWA testing adds to the predictive value of ejection fraction, thus further improving our ability to identify individuals whose risk for sudden death is elevated, particularly in combination with a noninvasive measure of autonomic function such as heart rate turbulence. Most directly relevant to the current CMS proposed decision memorandum, MMA-based TWA was found to be equivalent in predictive capacity to TWA measured by the spectral method.

T-wave alternans assessment by MMA as well as by the spectral method improves our ability to identify individuals who are at risk for sudden cardiac death. It is eminently clear that this parameter adds materially to information provided by ejection fraction determinations, particularly in patients with moderately depressed EF, in whom the sensitivity and specificity of EF alone is not sufficient to guide care.

In my opinion, it would be a backward step to remove reimbursement for TWA testing in general and specifically to decline to reimburse MMA, which has equivalent predictive power to the spectral method and which introduces the option to measure TWA from routine AECGs and exercise tests. As noted above, MMA’s predictivity has been demonstrated in sizeable prospective cohort studies. Therefore, I urge CMS to reconsider their decision to decline to cover reimbursement for MMA analysis of TWA.

Sincerely,

Peter H. Stone, M.D.
Associate Professor of Medicine, HMS
Co-Director of the Coronary Care Unit, Brigham & Women’s Hospital

Less

I am writing to register my comments regarding thepreliminary decision memo for T-wave alternans(TWA) testing with the Modified Moving Average(MMA) method. I am Director of the Non-invasiveElectrophysiology Unit at 1. Medizinische Klinikand Poliklinik, Technische Universität München. My interest in sudden death risk stratificationwith noninvasive methodologies spans more than twodecades.

1. Left ventricular ejection fraction, which hasbeen regarded as the gold standard in guiding ICDimplantation based on MADIT II criteria, has majorlimitations. Several studies have demonstrated,based on this parameter alone, that the ratio ofimplants to lives saved is in the range of 18:1. Furthermore, many individuals who die suddenlyhave preserved ejection fraction. Clearly, LVEFhas neither the sensitivity nor the specificity tobe the sole marker to identify individuals for ICDimplants.

2. Given the important role of the autonomicnervous system in precipitating sudden death, mylaboratory has been exploring the role ofautonomic markers, including heart rate turbulenceand deceleration capacity. Both of theseparameters, which can be monitored noninvasivelywith ambulatory ECG recorders, have proved to beequivalent or superior to ejection fraction,although they provide only indirect informationregarding cardiac electrical instability.

3. Thus, I maintain that TWA, which assessesrepolarization abnormalities, can contributeadditionally to evaluation of arrhythmia risk. The conventional approach to TWA analysis, thespectral method, is helpful in this regard, butbecause of its data stationarity requirements isnot suitable for ambulatory ECG monitoring. Because of the intrinsic flexibility of thetime-domain based MMA method, it is possible toassess TWA during both routine ambulatorymonitoring as well as exercise testing (Nieminenet al Eur Heart J 2007). This size of thedatabase (more than 1000 consecutive patients) ofthis prospective cohort study, its highlysignificant odds ratios in the range of 4 to 7,and the suitability of MMA for TWA analysis duringroutine exercise testing are impressive. Thislevel of enrollment and predictivity is comparableto the results achieved during an entire decadewith the spectral method.

4. In my opinion, it is in the context ofambulatory ECG monitoring that MMA-based TWAmeasurement provides its most significant advance. Particularly as TWA results can be combined withthe results of autonomic measures, including heartrate turbulence and deceleration capacity, ourability to estimate risk for sudden death isachieving unprecedented power. In a side-by-sidecomparison by Dr. Exner and coworkers (J Am CollCardiol 2007), MMA demonstrated comparablepredictivity to the spectral method. When TWAmeasurement was combined with autonomic testing,the results were superior. It is also worthnoting that these noninvasive measures aresuperior to LVEF alone.

Collectively, these studies show the value ofMMA-based TWA. Because of its demonstrated valuein clinical evaluation of patients, I believe thatit would be a step in the wrong direction toexclude MMA from reimbursement. Moreover, theprospect of reverting to ejection fraction as asingular parameter for sudden death riskstratification is not well founded and woulddeprive patients of the notable progress that hasbeen tested and found suitable in sizeableprospective cohorts.

PS: The computer system does not allow to enterGermany as state/country. In order to submit mycomment, I entered the first state on thepull-down menu (Alabama).

Less

I am submitting this comment on the proposed decision memo for Microvolt T-Wave Alternans (CAG-00293R). I concur with the proposed decision to continue National Coverage for Microvolt T-Wave Alternans (MTWA) for the evaluation of patients at risk for sudden cardiac death (SCD) only when the spectral analytic method is used and non-coverage when MTWA is measured by any other method. In particular, I agree with CMS's assessment that evidence is lacking to support demonstration of clinical efficacy of the modified moving average (MMA) method for measuring T-wave alternans.

I submitted a detailed public comment on the National Coverage Analysis on September 16, 2007. Since that time the Exner et al (1) study has been published and CMS in its proposed decision memo has requested comment on the implications of the MASTER 1 (2) trial which has not yet been published but was reported in oral form at the 2007 American Heart Association meeting. I will add to my earlier comments a discussion of these two items.

Exner et al Study (1)

The Exner et al study (1) (REFINE) enrolled 322 patients after myocardial infarction with left ventricular ejection fraction < 0.50. This study involved the use of multiple non-invasive diagnostic tests (including MMA and spectral analytic methods of measuring T-wave alternans) and reported multiple endpoints. The endpoint of greatest relevance to the MTWA National Coverage Decision is the one that refers to ventricular tachyarrhythmic events and does not include deaths due to non-arrhythmic causes. This endpoint is termed fatal or non-fatal cardiac arrest in this study. The study does not report T-wave alternans (measured using either method) as a univariate predictor of this endpoint. Instead, the study reports T-wave alternans only in combination with other diagnostic tests as a predictor of this endpoint. The study reports that the combination of T-wave alternans with other non- invasive diagnostic tests performed equivalently when T-wave alternans was measured by either method. However, it is not possible from the published paper to determine how each method of measuring T-wave alternans performed as a univariate predictor of fatal or non-fatal cardiac arrest. Interestingly, the study reported rather low correlation (kappa = 0.17) between the results of the MMA and spectral analytic methods.

Perhaps more importantly, in this study as in other studies, the cut-point for MMA T-wave alternans positivity was determined retrospectively from the analysis of the outcome data. Thus this study does not constitute a prospective evaluation of the MMA method. Furthermore, the specific cut-point value that was retrospectively determined is 5 microvolts which is more than an order of magnitude smaller than the 65 microvolt cut-point retrospectively determined in the larger Nieminen et al (3) study. The 5 microvolt cut-point retrospectively fit in the Exner et al study is also more than a factor of two less than the 10.75 microvolts fit retrospectively in the 41 patient Cox et al (4) study. Thus the MMA method currently lacks prospective trial data supporting its use because these studies (1, 3, 4) all fit the cut- point retrospectively. Moreover, the cut-points found in these retrospective analyses vary by a factor of 13. This very large inconsistency in retrospectively fitted cut-point values suggests the possibility that any association of MMA measured T-wave alternans with endpoint events could be spurious resulting from the retrospective fitting. Certainly these analyses do not establish a set of criteria which may be relied upon for interpretation of test results in clinical practice.

In contrast, the spectral analytic method of measuring MTWA has long standing peer reviewed published criteria (5) for clinical interpretation. These criteria have been applied prospectively in numerous prospective clinical trials which have demonstrated the predictive accuracy of MTWA when measured using the spectral analytic method. The equipment used to measure MTWA includes automated classification of test results based on these published criteria. The automated test result classification promotes uniform test result interpretation across medical practices. The MMA method has neither a set of accepted criteria for test interpretation nor an automated commercial algorithm for test interpretation.

MASTER 1 Trial

The MASTER 1 trial (2) was reported in oral form at the 2007 American Heart Association meeting. This study involved 575 patients with prior myocardial infarction and left ventricular ejection fraction ¡Ü 0.30. All of the patients had a spectral analytic MTWA test performed and all the patients had an ICD implanted. The primary endpoint was either arrhythmic death or appropriate ICD therapy. Patients were followed for a mean of 25 months. There were 70 endpoint events, 63 (90%) of which were appropriate ICD therapy. While the event rate was greater in the MTWA not-negative (positive or indeterminate) versus negative group (hazard ratio 1.26) this difference did not achieve statistical significance. MTWA was a significant predictor of total mortality (hazard ratio 2.04, p = 0.02).

The key confounding issue with the MASTER 1 trial is the fact that all the patients had implanted ICDs; accordingly the predominant arrhythmic endpoint was appropriate ICD therapy (here appropriate denotes that upon review of the electrogram immediately preceding the delivery of ICD therapy, the ICD was determined to have performed as it was intended to do). A number of investigators (6, 7) have reported that in ICD treatment trials that the number of patients receiving appropriate ICD therapy in the ICD arms of the trials exceed by a factor of 2 to 3 the number of sudden cardiac deaths in the treatment arm. The excess number of appropriate ICD therapies has been attributed to the fact that ICDs deliver therapy for arrhythmias that would have self-terminated had no ICD been implanted. An alternative explanation is that ICDs themselves may be arrhythmogenic and end up treating arrhythmias that they themselves induced (7). In any case the large excess number of patients receiving appropriate ICD therapy suggests that appropriate ICD therapy is a quite unreliable surrogate endpoint for sudden cardiac death in clinical trials.

In a published meta-analysis (8) of spectral analysis MTWA prospective trials involving primary prevention patients predominantly without implanted ICDs, similar to patients enrolled in MADIT-II (9) and SCD-HeFT (10), it was found that patients with a positive MTWA test had an annual rate of ventricular tachyarrhythmic events (VTEs) of 8.4% per year compared to an annual event rate of 1.2% per year for patients who tested MTWA negative yielding a hazard ratio of 7.0.

This same meta-analysis showed that in such trials that such patients, predominantly without implanted ICDs, who tested MTWA negative had an annual mortality rate of only 2.0% per year which is substantially lower than the 7.4% annual mortality rate observed in patients in MADIT-II and SCD-HeFT who did receive an ICD ¨C further indicating that a negative MTWA test confers a very low risk of sudden death even in patients comparable to those in MADIT-II and SCD- HeFT.

Since publication of this meta-analysis a number of additional large trials have been published and their results support the predictive accuracy of MTWA measured using the spectral analytic method in patients predominantly without implanted ICDs (e.g. Ikeda et al. (11), the non-ICD arm of Chow et al. (12), the ALPHA trial (13)). For example, the ALPHA trial (13) published in 2007, almost at the same time that the MASTER 1 trial was reported, involved 446 patients with non-ischemic dilated cardiomyopathy and left ventricular ejection fraction ¡Ü 0.40. This study reported that at one year of follow-up that 0.7% of the MTWA negative patients had sustained a VTE whereas 4.9% of the MTWA not-negative patients had sustained a VTE (hazard ratio 5.53, p = 0.004).

In contrast, a number of recent studies ¨C including MASTER 1 - have been presented involving patients with prophylactically implanted ICDs and involving appropriate ICD discharge as thepredominant arrhythmic endpoint.

[Formatted table: 'Studies Where the Predominant Arrhythmic Endpoint is Appropriate ICD Therapy' submitted via email attachment and available upon request.]

While some of these studies achieved statistical significance and some did not, in all of them the hazard ratio is substantially reduced compared to studies involving patients predominantly without implanted ICDs.

It appears that while MTWA measured using the spectral analytic method accurately predicts life threatening VTEs, it does not predict the occurrence of the excess appropriate ICD therapies which are not associated with life threatening arrhythmias. These excess appropriate ICD therapies when included as endpoints appear to function as statistical noise that randomly distribute among MTWA negative and not-negative patients, and thus obscure the predictive accuracy of MTWA for life threatening VTEs. Thus appropriate ICD therapy is an extremely unreliable surrogate endpoint for sudden cardiac death.

The data clearly and overwhelmingly demonstrate that MTWA measured using the spectral analytic method is an extremely accurate predictor of VTEs in patients who do not already have implanted ICDs. Conversely, there is no evidence demonstrating that ICD therapy is beneficial in primary prevention patients with a negative MTWA test measured using the spectral analytic method. In fact, a prospective non-randomized study (17) of 768 patients with ischemic cardiomyopathy (left ventricular ejection fraction ¡Ü 0.35) and no prior history of sustained ventricular arrhythmia found that ICD implantation was associated with a significant 55% reduction in mortality in the patients who had a not-negative spectral analytic MTWA test (p = 0.003), but observed no significant affect on mortality in patients with a negative test result. This study used the propensity score statistical methodology to adjust for factors that influenceICD implantation.

ICD therapy can be truly life saving in patients at risk. However, this therapy is not risk-free. The complication rate associated with just the ICD implantation procedure itself has been reported to be 11% including a mortality rate of 1% (18). The cumulative complication rate for ICDs has been found to be 31% over 46 months of follow-up (19). The DINAMIT trial (20) found that ICD therapy was associated with an increase in non-arrhythmic mortality of 2.6% per year (p = 0.02).

Thus ICD therapy may not benefit patients with a very low level of risk of sudden cardiac death as indicated by a negative spectral analytic MTWA test and could affect such patients adversely. MTWA testing using the spectral analytic method is a proven reliable tool to aid patients and physicians in guiding ICD therapy to at-risk patients who are likely to benefit from such therapy.

Sincerely,

Richard J. Cohen, M.D., Ph.D.
Whitaker Professor Harvard-MIT
Division of Health Sciences and Technology
Massachusetts Institute of Technology
Consultant and Board Member, Cambridge Heart, Inc.

References
1. Exner DV, Kavanagh KM, Slawnych MP, Mitchel BL, Ramadan D, Aggarwal SG, Noullet C, Schaik VA, Mitchell RT, Shibata MA, Gulamhussein S, McMeekin J, Tymchak W, Schnell G, Gillis AM, Sheldon RS, Fick GH, Duff HJ. Noninvasive Risk Assessment Early After a Myocardial Infarction. J Am Coll Cardiol. . 2007;50(24):2275- 2284.
2. Chow T, Kereiakes DJ, Onufer J, Woeffel A, Gursoy S, Peterson BJ, Brown ML, Pu W, Benditt DG. Late-Breaking Clinical Trials Abstract: Primary Results From the Microvolt T Wave Alternans Testing for Risk Stratification of Post MI Patients (MASTER I) Trial. Circulation. 2007;116(22):2631.
3. Nieminen T, Lehtimaki T, Viik J, Lehtinen R, Nikus K, Koobi T, Niemela K, Turjanmaa V, Kaiser W, Huhtala H, Verrier RL, Huikuri H, Kahonen M. T-wave alternans predicts mortality in a population undergoing a clinically indicated exercise test. Eur Heart J. 2007;28(19):2332-2337.
4. Cox V, Patel M, Kim J, Liu T, Sivaraman G, Narayan SM. Predicting arrhythmia-free survival using spectral and modified-moving average analyses of T-wave alternans. Pacing Clin Electrophysiol. 2007;30(3):352-358.
5. Bloomfield DM, Hohnloser SH, Cohen RJ. Interpretation and classification of microvolt T wave alternans tests. J Cardiovasc Electrophysiol. 2002;13(5):502-512.
6. Ellenbogen KA, Levine JH, Berger RD, Daubert JP, Winters SL, Greenstein E, Shalaby A, Schaechter A, Subacius H, Kadish A. Are implantable cardioverter defibrillator shocks a surrogate for sudden cardiac death in patients with nonischemic cardiomyopathy? Circulation. 2006;113(6):776-782.
7. Germano JJ, Reynolds M, Essebag V, Josephson ME. Frequency and causes of implantable cardioverter-defibrillator therapies: is device therapy proarrhythmic? Am J Cardiol. 2006;97(8):1255-1261.
8. Armoundas AA, Hohnloser SH, Ikeda T, Cohen RJ. Can microvolt T-wave alternans testing reduce unnecessary defibrillator implantation? Nat Clin Pract Cardiovasc Med. 2005;2(10):522-528.
9. Moss AJ, Zareba W, Hall WJ, Klein H, Wilber DJ, Cannom DS, Daubert JP, Higgins SL, Brown MW, Andrews ML. Prophylactic implantation of a defibrillator in patients with myocardial infarction and reduced ejection fraction. N Engl J Med. 2002;346(12):877- 883.
10. Bardy GH, Lee KL, Mark DB, Poole JE, Packer DL, Boineau R, Domanski M, Troutman C, Anderson J, Johnson G, McNulty SE, Clapp- Channing N, Davidson-Ray LD, Fraulo ES, Fishbein DP, Luceri RM, Ip JH. Amiodarone or an implantable cardioverterdefibrillator for congestive heart failure. N Engl J Med. 2005;352(3):225-237.
11. Ikeda T, Yoshino H, Sugi K, Tanno K, Shimizu H, Watanabe J, Kasamaki Y, Yoshida A, Kato T. Predictive value of microvolt T-wave alternans for sudden cardiac death in patients with preserved cardiac function after acute myocardial infarction: results of a collaborative cohort study. J Am Coll Cardiol. 2006;48(11):2268-2274.
12. Chow T, Kereiakes DJ, Bartone C, Booth T, Schloss EJ, Waller T, Chung ES, Menon S, Nallamothu BK, Chan PS. Prognostic utility of microvolt T-wave alternans in risk stratification of patients with ischemic cardiomyopathy. J Am Coll Cardiol. 2006;47(9):1820-1827.
13. Salerno-Uriarte JA, De Ferrari GM, Klersy C, Pedretti RF, Tritto M, Sallusti L, Libero L, Pettinati G, Molon G, Curnis A, Occhetta E, Morandi F, Ferrero P, Accardi F. 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(19):1896- 1904.
14. Costantini O, Rosenbaum DS, Hohnloser SH, Kirk M, Lerman B, Baker II J, Sethuraman B, Dettmer M. Late-Breaking Clinical Trials Abstract: The Alternans Before Cardioverter Defibrillator (ABCD) Trial: A Noninvasive Strategy for Primary Prevention of Sudden Cardiac Death Using T-Wave Alternans. Circulation. 2006;114(22):2426.
15. Gold MR, Ensley D, Chilson D, Ip JH, Costantini O, Bloomfield D, Poole JE, McNulty SE, Lee KL, Bardy G. Abstract 2113: T-Wave Alternans SCD HeFT Study: Primary Endpoint Analysis. Circulation. 2006;114:II_428- II_429.
16. Cantillon DJ, Stein KM, Markowitz SM, Mittal S, Shah BK, Morin DP, Zacks ES, Janik M, Ageno S, Mauer AC, Lerman BB, Iwai S. Predictive value of microvolt T-wave alternans in patients with left ventricular dysfunction. J Am Coll Cardiol. 2007;50(2):166-173.
17. Chow T, Kereiakes DJ, Bartone C, Booth T, Schloss EJ, Waller T, Chung E, Menon S, Nallamothu BK, Chan PS. Microvolt T-wave alternans identifies patients with ischemic cardiomyopathy who benefit from implantable cardioverter-defibrillator therapy. J Am Coll Cardiol. 2007;49(1):50-58.
18. Reynolds MR, Cohen DJ, Kugelmass AD, Brown PP, Becker ER, Culler SD, Simon AW. The frequency and incremental cost of major complications among medicare beneficiaries receiving implantable cardioverter- defibrillators. J Am Coll Cardiol. 2006;47(12):2493-2497.
19. Alter P, Waldhans S, Plachta E, Moosdorf R, Grimm W. Complications of implantable cardioverter defibrillator therapy in 440 consecutive patients. Pacing Clin Electrophysiol. 2005;28(9):926-932.
20. Hohnloser SH, Kuck KH, Dorian P, Roberts RS, Hampton JR, Hatala R, Fain E, Gent M, Connolly SJ. Prophylactic use of an implantable cardioverter-defibrillator after acute myocardial infarction. N Engl J Med. 2004;351 (24):2481-2488.

Less

This seems like a very biased read of the literature, especially when comparing the comments to that of the previous decision limiting the technique to the spectral analysis. For example, when discussing the Finish study, the critique that the study population is different than the CMS population wasn''t levied against a German study cited in support of the first CMS decision.

Most importantly, a study that appeared in JACC (Exner et al. Noninvasive risk assessment early after a myocardial infarction the REFINE study. J Am Coll Cardiol (2007) vol. 50 (24) pp. 2275-84) which directly compared the spectral method with the MMA method and found the identical predictive power between the two. This seems like the most compelling data to say that the techniques are likely measuring similar things.

Less