We agree with Drs. Carbajal, Huang, and Hu that several trials included in our systematic review had potential methodologic deficiencies. Although we had space to highlight only 1 of these deficiencies in our published manuscript (randomization after implantation of the CRT device in all but 1 trial), we refer interested readers to our full Agency for Healthcare Research and Quality report, available at http://www.ahrq.gov/clinic/tp/resyntp.htm, for a full description of the methods and the quality assessment tables for all of the studies included in both our efficacy and safety analyses.

We refute the allegation of Drs. Calvert, Freemantle, and Cleland that our analysis of the COMPANION trial data (1) "ignores a basic tenet of meta-analysis." As outlined in our manuscript, we did conduct intention-to-treat analyses for all end points, and halving the control group data when incorporating a 3-arm trial into a meta-analysis is one of the techniques endorsed by the Cochrane Collaboration, as outlined at http://www.cochrane-net.org/openlearning/HTML/modA2-5.htm. Indeed, we would like to point out that, despite Calvert and colleagues' assertion that we created incorrect decreases for the CRT-alone arm of the COMPANION trial, careful perusal of our Figure 2 reveals that the relative risk we reported for mortality in that arm (0.84 [CI, 0.61 to 1.14]) is not statistically significant and is in fact very similar to the relative risk of 0.85 (CI, 0.67 to 1.09) that they found in their analysis.

On the other hand, Dr. Calvert and colleagues are correct in pointing out (as did Drs. Hlatky and Massie in the editorial accompanying our paper [2]) that exclusion of the CRT-ICD arm of the COMPANION trial would change the pooled relative risk for mortality with CRT to 0.84 (CI, 0.69 to 1.03). We do not disagree that this is one way to approach the data, but we wonder to what extent this argument is moot in light of the current ICD evidence base, particularly given the data from the Sudden Cardiac Death in Heart Failure Trial (SCD-HeFT) (3, 4). Shouldn't ICDs be considered in patients who have substantial heart failure symptoms despite medical therapy, prolonged QRS duration, and a left ventricular ejection fraction less than 0.35 (the base case in our systematic review and the subsequent decision analysis), unless there are contraindications or competing comorbid conditions? Indeed, we believe that the vast majority of patients with New York Heart Association class III symptoms who are being evaluated for CRT will also be eligible for an ICD. Given this, aren't the data from the CRT-ICD arm of the COMPANION trial relevant to the discussion? Furthermore, it should be emphasized that the beneficial effects of CRT on functional status, left ventricular ejection fraction, and quality of life reported in our study are present even without inclusion of the COMPANION CRT-ICD data.

Dr. Calvert and colleagues state that the heart failure hospitalization data are inconclusive. Certainly, this is true of the pooled estimate that we reported and that they reiterate in their letter. However, in reexamining our data, we discovered an error in Figure 3 in our published manuscript. We mistyped the number of "no deaths" in the CRT group for the Guidant CONTAK CD CRT-D System Trial (5) when entering the data into our models. A corrected Figure is included here (Figure). The corrected pooled relative risk for heart failure hospitalizations is thus 0.67 (CI, 0.48 to 0.92), a result that is statistically and clinically significant and does not include the COMPANION trial results (the data for this end point were not available at the time of our request to the COMPANION investigators).

View larger version (16K): [in this window] [in a new window]   Figure. Heart failure hospitalizations with cardiac resynchronization therapy (CRT) versus controls. CHF = congestive heart failure; CONTAK-CD = Guidant CONTAK CD CRT-D System Trial; MIRACLE = Multicenter InSync Randomized Clinical Evaluation; MIRACLE-ICD = Multicenter InSync Randomized Clinical Evaluation ICD [implantable cardioverter defibrillator]; MUSTIC-AF = Multisite Stimulation in Cardiomyopathies Atrial Fibrillation; MUSTIC-SR = Multisite Stimulation in Cardiomyopathies Sinus Rhythm; RR = relative risk.

While we disagree with the specific points that Dr. Calvert and colleagues raise in the first half of their letter, we do not disagree that the trials in our analysis may have overestimated the benefits of CRT. Indeed, we made this very point in our manuscript. We also agree with these CARE-HF investigators that our meta-analysis (which is based on only 429 deaths and 251 heart failure hospitalizations) should not be used as a reason to prematurely terminate ongoing large trials in this area (such as CARE-HF and RAFT [Resynchronization/Defibrillation for Advanced Heart Failure Trial]).

In closing, we would like to reiterate 3 key points from our original study that we fear some readers may have missed: 1) CRT is not a panacea; 2) CRT is efficacious in selected patients with advanced systolic heart failure and evidence of electromechanical dyssynchrony; and 3) attempts to define which patients are most likely to benefit from CRT should remain a research priority.