Previous studies have suggest that the authors of randomized trial reports frequently use rhetorical strategies, such as framing, the use of ratios rather than absolute values to report results, and underreporting of complications, to convince readers of treatment efficacy. The objective of this study was to determine if such rhetorical strategies were used in the publication of implantable-cardioverter defibrillator (ICD) primary prevention trials. Medline and the Cochrane Central Register of Controlled Trials were searched for all publications that described ICD primary prevention trials and that involved >100 subjects. Each publication was analyzed for evidence of message framing, the exclusive use of ratios to report outcomes, underreporting of ICD complications, and interpretation bias favoring ICD therapy. Ten publications were identified. Introductory remarks in 8 of the 10 publications cited the evaluation of ICD benefits as the sole objective and mentioned only background studies that supported ICD efficacy, suggesting message framing in support of ICD efficacy. Five publications provided no specific information about the frequency of unsuccessful ICD implantations, and 8 publications provided incomplete or no information about implantation and postimplantation complications. ICD complications were not mentioned in the discussion sections of 9 publications, and none of the publications included a comprehensive comparison of ICD benefits versus risks, consistent with interpretation bias. Ratios and statistically insignificant data were not used to emphasize ICD benefits. In conclusion, message framing, underreporting of ICD complications, and interpretation bias were used to emphasize ICD efficacy in the reporting of ICD primary prevention trials.
The American College of Cardiology, American Heart Association, and Heart Rhythm Society guidelines for the treatment of heart failure recommend implantable-cardioverter defibrillator (ICD) therapy for the primary prevention of sudden cardiac death in patients with heart failure, left ventricular ejection fractions ≤35%, and New York Heart Association functional class II or III symptoms. Nevertheless, the decision to implant a prophylactic defibrillator in a particular patient is often complicated, because most potential candidates will never experience life-threatening tachyarrhythmias, and the risks, such as pneumothorax, infection, lead problems, and inappropriate ICD shocks, are substantial. The medical community’s view of prophylactic ICD benefits and risks is heavily influenced by prophylactic ICD trial reports. Previous studies suggest that the authors of randomized trial reports often use rhetorical strategies to convince readers of treatment efficacy. Background information that does not support treatment efficacy often is not mentioned. Ratios rather than absolute values are sometimes used to report outcomes, because they give the appearance of greater treatment effects. Complications are underreported. Treatment benefits are frequently highlighted even in trials with statistically nonsignificant results. The present study was undertaken to determine if such rhetorical strategies were used in primary prevention ICD trial reports.
Methods
Reports of ICD primary prevention trials were identified by searching Medline, the Cochrane Central Register of Controlled Trials, the Cochrane Library, and previous reviews and meta-analyses of ICD primary prevention trials. The abstracts of all retrieved reports were reviewed to identify publications with the following characteristics: (1) the trial was designed to evaluate ICD implantation in patients with no histories of cardiac arrest or sustained ventricular tachyarrhythmias; (2) the publication was the first and primary report of trial results; (3) the trial included a study population of ≥100 patients; and (4) the trial did not involve cardiac resynchronization therapy.
Publications were analyzed for the presence of 4 common rhetorical strategies that can be used to emphasize treatment efficacy: message framing, the use of ratios rather than absolute values to report results, underreporting of harms, and interpretation bias. Framing consists of shaping perceptions to support a conclusion. The introduction section of a trial report can be used to frame expectations of a positive treatment effect. This can be done by stating that the trial objective is to determine treatment efficacy without mentioning the evaluation of treatment harms and by mentioning only background information that supports treatment efficacy. Ratios tend to magnify apparent treatment benefits. For example, a treatment that reduces mortality from 20% to 10%, indicating a 10% absolute decrease in mortality, can be said to decrease mortality by 50%, a far more impressive effect. Underreporting of important treatment complications observed during a trial will give the treatment a more favorable benefit to risk profile than warranted by trial results. The discussion section of a trial report can provide a misleading impression of trial results if important trial limitations and treatment complications are not mentioned and addressed and if conclusions are drawn from the trial that are not supported by trial results.
In ICD publications, the following presentation structures were considered evidence of message framing: (1) in the abstract and/or introduction, the objective of the trial was stated to be the evaluation of ICD benefits, with no mention of evaluating ICD risks, and (2) in the introduction, the only background information provided supported ICD efficacy.
The exclusive use of ratios to report primary outcome results was considered evidence that ratios were used to emphasize ICD benefits.
Complication data were considered underreported if (1) ICD complications were not mentioned in the abstract when significant ICD complications were observed during the trial, and/or (2) the results section did not contain information about the rate of unsuccessful device implantation, the types and frequency of procedural complications, the types and frequency of postimplantation complications, and/or the frequency of inappropriate ICD shocks.
The following presentation characteristics were considered evidence of interpretative bias: (1) important trial limitations were not mentioned and discussed; (2) ICD complications were not discussed, despite the presence of significant device complications in the study population; (3) the risks and benefits of ICD therapy were not compared, including how the frequency of risks should alter the use of ICDs; (4) the probability that ICD benefits are likely to be less in clinical practice than in a clinical trial was not mentioned; and/or (5) trial conclusions indicated that ICD therapy is unequivocally beneficial despite a failure of the trial to demonstrate a significant effect on the primary efficacy end point.
Results
Ten publications met the criteria set for inclusion in the study. These publications are listed in Table 1 . The results of the publication analysis are listed in Table 2 .
| Trial (Sponsor) | Comparison | n | Inclusion Criteria | Primary End Point | End Point Achieved |
|---|---|---|---|---|---|
| MADIT | ICD vs medical therapy (antiarrhythmic drugs in 92%) | 196 | Age 25–80 years, previous MI, NYHA I–III, EF ≤35%, NSVT, nonsuppressible ventricular tachyarrhythmia on EP testing | All-cause death | Yes |
| CABG Patch | ICD vs medical therapy | 900 | Age <80 years, CAD, EF <36%, elective CABG, abnormal results on SA ECG | All-cause death | No |
| MUSTT ⁎ | EP guided therapy (±ICD) vs medical therapy | 704 | Age <80 years, CAD, EF ≤40%, NSVT with inducible sustained VT/VF | Cardiac arrest or arrhythmic death | Yes |
| MADIT II | ICD vs medical therapy | 1,232 | Age >21 years, CAD, EF <30%, MI ≥1 month before entry | All-cause death | Yes |
| CAT | ICD vs medical therapy | 104 | Age 18–70 years, DCM ≤9 months, NYHA class II or III, EF ≤30% | Mortality at 1 year | No |
| AMIOVIRT | ICD vs amiodarone | 103 | Age ≥18 years, DCM, NYHA class I–III, EF ≤35%, asymptomatic NSVT | All-cause death | No |
| DEFINITE | ICD vs medical therapy | 458 | DCM, NYHA class II or III, EF <36%, NSVT or >10 PVCs/hour | All-cause death | No |
| DINAMIT | ICD vs medical therapy | 674 | Age 18–80 years, CAD, NYHA class III or IV, EF ≤35%, MI 6–40 days previously, impaired cardiac autonomic function | All-cause death | No |
| SCD-HeFT | ICD vs amiodarone vs medical therapy | 2,521 | Age ≥18 years, CAD or DCM, NYHA class II or III, EF ≤35% | All-cause death | Yes |
| IRIS | ICD vs medical therapy | 898 | Age 18–80 years, CAD, EF ≤40%, acute MI with HR ≥90 beats/min on first ECG and/or NSVT | All-cause death | No |
⁎ The MUSTT trial was primarily a trial of EP-guided therapy, including the use of an ICD, rather than an ICD primary prevention trial.
| MADIT | CABG Patch | MUSTT | MADIT II | CAT | AMIOVERT | DEFINITE | DINAMIT | SCD-HeFT | IRIS | |
|---|---|---|---|---|---|---|---|---|---|---|
| Message framing | ||||||||||
| Abstract study objective did not include evaluation of harms | • | • | • | • | • | • | • | • | • | • |
| Introduction study objective did not include evaluation of harms | • | • | • | • | • | • | • | • | • | • |
| Background information only supported ICD therapy | • | • | • | • | • | • | • | • | ||
| Ratio use | ||||||||||
| Primary end point results reported in the abstract only using hazard ratios | • | • | • | • | ||||||
| Primary end point results reported in the results section only using hazard ratios | ||||||||||
| Underreporting of complications | ||||||||||
| ICD complications not mentioned in the abstract | • | • | • | • | • | • | • | • | • | • |
| Frequency of unsuccessful device implantation not reported | • | • | • | • | • | |||||
| Number and types of complications during device implant not fully reported | • | • | • | • | • | • | • | • | ||
| Number and types of device complications after implant not fully reported | • | • | • | • | • | • | • | • | ||
| Frequency of inappropriate ICD shocks not clearly reported | • | • | • | • | • | • | • | • | • | |
| Important trial limitations not mentioned and discussed | • | • | ||||||||
| Interpretive bias | ||||||||||
| Complications not mentioned in the discussion section | • | • | • | • | • | • | • | • | • | |
| No comparison of risks vs benefits and how risks should alter ICD use | • | • | • | • | • | • | • | • | • | • |
| No comment that device benefits are likely to be less in clinical practice | • | • | • | • | • | • | • | • | • | • |
| Conclusions encourage ICD use despite failure to achieve the primary end point | ||||||||||
| Total (15 possible) | 10 | 12 | 12 | 11 | 9 | 11 | 7 | 11 | 9 | 11 |
Introductory remarks strongly suggest that message framing was used to emphasize ICD efficacy. In all the publications, the only trial objective noted in the introduction is the evaluation of ICD efficacy, with no mention of evaluating ICD risks. The background information reviewed in the introductory remarks typically makes the case for ICD therapy by emphasizing the risks of sudden death in the target population and by referencing previous trials that demonstrated an impact of ICD therapy on cardiac mortality. There are only 2 comments about negative trials, 1 in the Cardiomyopathy Trial (CAT) introduction about the negative results of the Coronary Artery Bypass Graft Patch (CABG Patch) trial and 1 in the Immediate Risk-Stratification Improves Survival (IRIS) introduction about the negative results of the Defibrillator in Acute Myocardial Infarction Trial (DINAMIT) trial. None of the introductory sections mention ICD risks as a potential problem or cite the incidence of ICD complications in previous studies.
There was no evidence that ratios and statistically insignificant data were used to emphasize ICD benefits. The effect of ICD therapy on survival is exclusively reported using hazard ratios in the Multicenter Automatic Defibrillator Implantation Trial (MADIT), CABG Patch, DINAMIT, and IRIS abstracts. However, none of the publications relies exclusively on ratio measures to report primary outcomes data in the results sections. All publications display primary outcome data using Kaplan-Meier plots.
Underreporting of ICD complications occurred frequently. ICD complication data reported in the 10 publications are listed in Table 3 . Five publications provide no specific information about the frequency of unsuccessful ICD implantations. Eight publications provide incomplete or no information about the frequency of implantation and postimplantation complications. Nine publications do not report the frequency of inappropriate ICD shocks.
