Can’t get out of this world alive Free

This editorial refers to ‘Reasons for hospitalization and risk of mortality in patients with atrial fibrillation treated with dabigatran or warfarin in the Randomized Evaluation of Long-term Anticoagulation Therapy (RE-LY) trial’, by A. Alak et al., on pages 1023–1030.

In this issue of EP-Europace, Alak et al. report a post hoc analysis of the RE-LY Trial in which they consider reasons for hospitalization and risk of mortality in patients with atrial fibrillation (AF) treated with dabigatran or warfarin.¹ Of 18 113 patients, 7200 (39.8%) were hospitalized at least once during a mean follow-up of 2 years with similar rates of hospitalizations for those taking dabigatran and those taking warfarin. Hospitalization was associated with a higher likelihood of post-discharge death [absolute event rate 9.1% vs. 2.2%, adjusted hazard ratio (HR) 3.6, 95% confidence interval 2.4–3.3; P-value <0.0001] as well as increased risk of vascular and sudden cardiac death (similar HRs and P-values). Cardiovascular hospitalization was likewise associated with increased risk of post-discharge death, vascular death, and sudden death (similar HRs and P-values). The authors assert: ‘… hospitalization and cardiovascular hospitalization, if clearly pre-specified and defined, may serve as a surrogate for all-cause mortality in AF trials’. Really?

Let’s consider what this study shows: (i) hospitalization (cardiovascular or otherwise) was associated with increased risk of post-discharge death compared to those not hospitalized whether or not the hospitalization was related to AF. (ii) All patients had AF.

Why hospitalize?

People are hospitalized because they are sick, regardless of whether they have AF or not. Indeed, sicker people are more likely to die sooner despite hospitalizations that may not cure the problem but, hopefully, improve, or at least stabilize, the underlying process.

What does atrial fibrillation have to do with hospitalization…or death?

Atrial fibrillation may have nothing to do with most hospitalizations or the outcome. In this analysis, the strength of the association was greater with paroxysmal rather than persistent AF. Something else is at play here. Fifty-nine percent of hospitalizations were non-cardiovascular and cardiovascular hospitalizations were not necessarily AF-related. Thirty percent, or so, had cardiovascular hospitalizations related to AF (but not necessarily the primary diagnosis). Thirteen percent of all hospitalizations were AF-related. Furthermore, AF may have had nothing to do with the fact that patients died.

Why consider a surrogate endpoint?

A surrogate endpoint is a reliable substitute, supported by a clear, mechanistic, rationale, that may have value if the original endpoint requires an extended period to achieve or is otherwise difficult to attain.^2,3 The surrogate endpoint should be an accurate, robust, and specific predictor. Nearly 40% were hospitalized during the short follow-up. In this case, hospitalization was neither a specific, sensitive, or robust marker for death. Death was not shown to be related to AF. In this report, the value of hospitalization as a surrogate was not demonstrated by any time savings, so, there was no purported value of substituting one endpoint for another except, possibly, to power a trial but, at the same time, pollute it.

Is hospitalization a robust marker for death?

In AF trials, consideration of hospitalization as a surrogate marker for mortality remains unproven. Hospitalization in this trial did not predict mortality with any degree of certainty and was not shown to predict mortality related to AF. While the HRs showed that hospitalization was associated with greater risk of dying for those hospitalized for any reason but the absolute event rate was only 9.1% meaning that 90.9% of those hospitalized did not die during follow-up. Indeed, this risk is statistically higher than 2.1% (for those not yet hospitalized) but, as a surrogate marker for death, it is not robust and, therefore, there are concerns and dangers using this as a surrogate. Even if the HR is high, and the P-value is highly significant, most hospitalized patients do not die or otherwise differ from those who are not hospitalized.

Can one correct for baseline comorbidities?

The sicker the patient, the more likely the patient will be admitted. The authors attempt to ‘correct’ the data for covariate predictors but, in this post hoc study, even with a multivariate analysis, one cannot equate hospitalized patients to those who were not hospitalized. If one could predict hospitalization from baseline comorbidities, perhaps, there would be no need to reassess a patient in any specific situation. A post hoc multivariate analysis of comorbidities does not get to the truth of the matter in this study.

Was a temporal relationship between hospitalization and death found?

If hospitalization were associated with, or followed immediately by, death, hospitalization could not be considered helpful, and may even be harmful, but the present report does not provide the temporal relationship between hospitalization and death. If a long-time interval existed between hospitalization and death, one could conclude that a sick patient who was hospitalized had an extension of life due to hospitalization, yet, still died. Time to death is a critical measure as death will eventually reach us all. It’s just a matter of time. As was said in an old blues song about imbibing ethanol (‘I’m Drinking Again’, by Ry Cooder): ‘The doctor said it’d kill me but he didn’t say when’.

Economic burden of hospitalization

The authors consider hospitalization an economic burden and, while this may be the case, they offer no alternative. Not hospitalizing a sick patient? Without a carefully controlled, prospective trial evaluating this issue, no conclusion can be drawn. While iatrogenic misadventures during hospitalization may play a role in a post-discharge death, I cannot imagine a study that will randomize a sick patient in need of hospitalization to hospitalization or no hospitalization to find this out. An alternative solution: Since older patients, and those with more comorbidities, are more likely to be hospitalized, and die sooner rather than later, one solution would be to encourage an earlier death, by some undefined means. It could save costs and banish needless suffering!

The conclusion

Alak et al.¹ conclude: ‘[These findings]…emphasize the importance of preventing hospitalization among AF patients and adequate surveillance post-discharge’. Adequate surveillance is generally a great idea but there is no proposal on how to prevent hospitalizations. No data indicate any means to prevent hospitalizations, and no data indicate any value of surveillance post-discharge. The impact of not hospitalizing a sick patient may be far, far worse.

Alak et al.¹ state: ‘These findings may support the use of hospitalization and/or cardiovascular hospitalization as surrogate outcomes in future AF studies’. I respectfully disagree.

‘No matter how I struggle and strive—I’ll never get out of this world alive’—Hank Williams, American country singer, died at age 29, in 1953 in Oak Hill, West Virginia at the time these lyrics to his #1 tune were being aired. He had the temerity to die suddenly without even being hospitalized.

Conflict of interest: B.O. reports the following conflicts: Boehringer Ingelheim (co-national director GLORIA AF), Amarin (chair of DSMB of REDUCE-IT), Lundbeck (speaker and consultant), Respironics (consultant), and Sanofi Aventis (speaker).

The opinions expressed in this article are not necessarily those of the Editors of Europace or of the European Society of Cardiology.

References