When is an implantable Cardioverter-Defibrillator Controversial?

There are few areas of medicine that have no controversial aspects. It has been argued that more than 95% of medical practice is based on expert opinion rather than guideline recommendations based on adequately powered randomised controlled trial (RCT) evidence. In heart failure, many of our standard treatments are based on such evidence, and it is arguably the area of medicine most replete with such an evidence base for so much of our routine clinical care. The exception, I hear you all say, is our use of diuretics, which almost all heart failure patients need at some stage, but which is almost entirely free of RCTs to guide how and when to use these agents. There is one area of heart failure practice where we all felt the evidence had settled and we knew what to do, when and to whom, and that is the relatively precise treatment option of an implantable cardioverter-defibrillator (ICD), as a treatment to defibrillate a patient undergoing ventricular tachyarrhythmic-threatened sudden death, even if it remained a topic of interest for further study.1-4 In this area of practice, most major guidelines until recently have shown a fair degree of consensus concerning when ICDs should be implanted for primary or secondary prevention. The aim of primary prevention of course is to prevent arrhythmic sudden death in someone who has not suffered such a threatened event in the past, and secondary prevention is the use of such a device in someone who has recovered or has been resuscitated from such a threatened sudden death. In the 2016 European Society of Cardiology guidelines,5 it was stated that for primary prevention in a patients with heart failure with reduced ejection fraction (HFrEF) ‘an ICD is recommended to reduce the risk of sudden death and all-cause mortality in patients with symptomatic heart failure (New York Heart Association class II–III), and a left ventricular ejection fraction ≤ 35% despite ≥ 3 months of optimal medical therapy, provided they are expected to survive substantially longer than 1 year with good functional status, and they have: (i) ischaemic heart disease (unless they have had a myocardial infarction in the prior 40 days); (ii) dilated cardiomyopathy’. It was given a class 1A recommendation for both ischaemic and non-ischaemic heart failure (NIHF). At this time, few contemporaneous guidelines differed with this view, despite the fact there had not been a definitive dedicated trial of ICD insertion in NIHF. What was the evidence base in 2016? Two unconfounded trials of the use of ICDs in entirely non-ischaemic HFrEF population had reported. One (CAT trial) was terminated early, because originally planning to recruit 1348 patients with an expected 1 year mortality rate of 30% in the first year, it was terminated early with only 104 patients and a miserly mortality rate of 5.6% overall.6 The other (DEFINITE)7 was powered on the assumption of 2-year mortality rates of 15% in the standard-therapy group and 7.5% in the ICD group. This is therefore planning and expecting a 50% reduction in total mortality to be valid. Given that we know ICDs are only ever going to reduce sudden arrhythmic deaths and that such a mode of death is in the minority in all major heart failure trials (see later), it was, one may argue, doomed to fail by being underpowered. It simply could not reduce more deaths than were likely to treatable by ICD therapy. In the end, DEFINITE recruited 458 NIHF patients and showed a strong trend to benefit for the impossible to achieve the primary endpoint (for a trial of this size) of all-cause mortality. There were 68 deaths: 28 in the ICD group, as compared with 40 in the standard-therapy group [hazard ratio (HR) 0.65; 95% confidence interval (CI) 0.40–1.06; P = 0.08]. The only true target for an ICD (i.e. sudden deaths from arrhythmia) was very significantly reduced with 3 in the ICD group, as compared with 14 in the standard-therapy group (HR 0.20; 95% CI 0.06–0.71; P = 0.006). There was more evidence, however, for ICDs in NIHF, and that is because NIHF was a large subgroup in the overall positive ICD trial in HFrEF, SCD-HeFT.8 It was sub-group, but an important and pre-specified one, and there was no suggestion that the non-ischaemic patients had a worse response to ICD insertion. It is a principle of interpreting trials that we should be conservative when assessing a sub-group behaves differently to the rest of the trial. One should make the starting assumption that all sub-groups are the same as the overall trial population and only move away from that in very specific circumstances. It would require a statistically significant interaction term, a large difference in efficacy that is clinically meaningful and a mechanism for the effect that is biological plausible. By this benchmark we should not assume the NIHF patients behaved any differently to the overall trial results of SCD-HeFT. It would be the equivalent of saying we should not recommend angiotensin-converting enzyme inhibitors in women with HFrEF because there have been no trials in women alone and the sub-group analyses of women by themselves are not separately significant. Lastly, there was a well-conducted meta-analysis9 that concluded for all-cause mortality: ‘ICD benefit was similar for coronary artery disease (relative risk 0.67; 95% CI 0.51–0.88) vs. dilated cardiomyopathy (relative risk 0.74; 95% CI 0.59–0.93).’ Has anything come up to change our opinions? Since the publication of all these guidelines, the DANISH trial was published in September 2016.10 The trial was neutral, because its primary endpoint, all-cause mortality, was not significantly reduced, even though the hazard ratio was usefully below 1.0 (actually 0.87; 95% CI 0.68–1.12; P = 0.28). In the same trial the risk of sudden death was fairly dramatically and significantly reduced (4.3% vs. 8.2%; HR 0.50, 95% CI 0.31–0.82; P = 0.005). How should we in fact respond to a neutral trial? This is not an easy question to answer, because not proving a positive is not the same as proving a negative, it may be simply that the trial was underpowered or that we were expecting too much of the treatment. Certainly, there was no suggestion of any adverse effect on the primary outcome. On a purely statistical basis had this trial been the same size as PARADIGM-HF and had the point estimate of efficacy stayed the same, it would have been a clearly positive trial. Therefore, it is a case of ‘not proven’, rather than ‘guilty’. Since DANISH, several guideline updates have been published. Canadian guidelines11 maintained no differentiation between ischaemic and non-ischaemic heart failure by recommending an ICD with a ‘strong recommendation, high quality of evidence’ classification, and the American Heart Association/American College of Cardiology/Heart Rhythm Society update12 has maintained their class IA recommendation for primary prevention ICD in patients with class II–III NIHF and left ventricular ejection fraction ≤ 35%. The Australian guidelines published in August 2018, however, downgraded their recommendation or ICDs in NIHF to a ‘weak recommendation, low level of evidence’ in response to DANISH,13 although it is noted this has received local criticism as being a premature response.14 Thus, the eventual outcome of DANISH meant their pre-trial power assumptions were off the mark. In DANISH there were 131 deaths (23.4%) in the control group, in 95 this was cardiovascular (i.e. 27% non-cardiovascular) and 46 were sudden cardiac deaths; i.e. 35% of deaths overall were potentially salvageable by ICD. This is one-fourth lower than their estimates for the proportion of sudden deaths based on their stated sources of data for power calculations. This would not matter if there had been reserve power or if the mode of death did not matter, but for ICD there is a clear specificity for which types of death it can prevent and it is arguable that power calculations should only have been performed with that mode of death in mind. In addition, the death rate was lower and the trial leadership extended follow-up to compensate and it is not clear that sudden death rates do not fall with time even further in NIHF. Two critical questions arise from DANISH: (i) is it at all reasonable or even sensible to design a primary endpoint of all-cause mortality as a fair test, given the only reasonable mode of action of ICD is to reduce arrhythmic sudden death, the frequency of which often is lower than in the past; and (ii) what do we do with neutral studies after a consensus has already been reached? The DANISH trial result can be considered as one of testing a therapy in a population with too few treatable events (sudden deaths) to prove a potentially effective treatment against the background of a larger number of untreatable (for this particular therapy) events. If this is the case, then one should view this as an issue of number needed to treat to save lives which depends crucially on the prior risk of sudden death in the population in question. Guidelines have been criticised as often being a listing of treatment trials that have reduced total mortality, as if nothing else in medicine matters. Surely, practising cardiologists also want to know about and be recommended treatments that are extremely effective at preventing sudden arrhythmic death, even if that cannot with certainty be proven to affect overall mortality due to this particular mode of death being in the minority in contemporary populations. Perhaps the guidelines should say ICDs have a class 1A recommendation to prevent arrhythmic sudden death and leave physicians to choose patients with a need to reduce that mode of death in particular. Physicians are used to applying recommendations to patients who are not precisely those studied in the trials upon which the guidelines were based.18 It is thus interesting that the Heart Failure Association in its recent guidance update19 stated on this matter an unusual recommendation. After reiterating ICD should be used in HFrEF, it then proferred the qualifier that ‘one may consider not to implant an ICD in patients with non-ischaemic HFrEF who (i) are aged > 70 years, or (ii) have advanced symptoms (New York Heart Association class III/IV), or (iii) have life-shortening co-morbidity (e.g. severe lung disease or Stage IV chronic kidney disease) and hence are likely to die for reasons other than sudden arrhythmic death.’ Of course many factors should be taken into account in clinical decision making, many of which will not be based entirely on RCTs designed specifically to answer that question. For example, sometimes the best we have is retrospective sub-group analyses, such as for instance the recent report that suggested primary preventive ICD might be less effective in the presence of diabetes compared to its absence.20 This brings back clinical judgement into guidelines; a trend that I believe many physicians would applaud. Conflict of interest: none declared.

Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.