This editorial refers to ‘Cardiovascular hospitalization as a surrogate endpoint for mortality in studies of atrial fibrillation: report from the Stockholm Cohort Study of Atrial Fibrillation’ by L. Friberg and M. Rosenqvist, on page 626.
Cardiovascular (CV) hospitalization has been used as a primary endpoint in clinical studies on atrial fibrillation (AF), but its value remains controversial. Cardiovascular hospitalization is associated with a decreased quality of life,1 increased costs of care,2 and increased risk of mortality.3,4 However, it is also composed of a multiplicity of different clinical conditions ranging from heart failure to angina, and rates may vary among geographical regions and clinical settings. The operational definition and value of CV hospitalization as an endpoint should be well understood to draw valid conclusions in AF research.
Mortality is a relevant outcome to assess the effect of therapies in AF patients. However, because death rates are low and existing therapies are effective, mortality trials are challenging to conduct. From the feasibility and cost perspective it is useful to employ surrogate and combined endpoints, but these should be well understood and validated.5,6 Cardiovascular hospitalization is a potentially useful outcome of clinical importance. The Atrial Fibrillation Follow up Investigation of Rhythm Management (AFFIRM) investigators have shown that CV hospitalization could be a valid outcome in AF trials since it occurs prior to and more often than the true endpoint (mortality), can be objectively measured, and accurately predicted mortality independent of the application of rate or rhythm control.3 The first time CV hospitalization or death was used as a primary outcome in a major AF trial was in the ATHENA (A placebo controlled, double blind Trial to assess the efficacy of dronedarone for the prevention of cardiovascular Hospitalization or death from any cause in patiENts with Atrial fibrillation and flutter) study, which was also the first study to demonstrate a reduction in important clinical CV outcomes with an antiarrhythmic drug (dronedarone).7
With the Stockholm Cohort Study of Atrial Fibrillation (SCAF) analysis, Friberg and Rosenqvist4 demonstrate an association between CV hospitalization and all-cause mortality among 2912 AF patients during a 6.5-year follow-up using a prospective observational cohort with detailed patient information. The SCAF analysis confirms the original findings of the AFFIRM analysis regarding the significant association of CV hospitalization with mortality. The strength of the association did not differ greatly from that in AFFIRM, although it depends on which SCAF analysis you look at. The AFFIRM reported a hazard ratio of 2.15 in the rhythm control group and 1.71 in the rate control group.3 The SCAF hazard ratio for mortality during the first 3 months following hospitalization was 1.69, but decreased to 1.43 when taking into account deaths during 1 year and even to 1.35 when using a propensity score to correct for unmeasured differences between groups and improve comparability in this observational study. When adding medication use to the model of the final 2.5 years of the follow-up, the hazard ratios for mortality during 3 and 12 months were 2.08 and 1.63 respectively, which compared favourably with those observed in AFFIRM. All analyses of the SCAF verified the significance of the association of CV hospitalization and mortality, but also indicated the importance of the nature of the analysis to assess the strength of the association.
Using hospitalization as an endpoint is not a new concept since hospitalization for heart failure has been widely used in heart failure research. Although any CV hospitalization is less of a disease-specific endpoint in AF research, its use as a mortality substitute seems justified considering that a two-fold increased risk for mortality associated with AF mainly relates to the presence of concomitant CV disease and risk factors.8 The exact definition of CV hospitalization therefore needs some thought. The SCAF analysis showed that hospitalization primarily for AF was associated with lower mortality than hospitalization for other CV diagnoses. Considering that AF hospitalization was the main driver for CV hospitalization, with 44% of all hospitalizations, a treatment specifically targeting AF might lower the overall CV hospitalization rate without significantly affecting mortality. In the ATHENA study the reduction in the primary outcome with dronedarone was mainly due to a 37% relative risk reduction in AF hospitalizations, but there was also a trend towards lower all-cause mortality.7 The latter might relate to the drug's effect on AF hospitalizations, the concomitant reductions in hospitalizations for acute coronary syndromes and heart failure, or the fact that only unplanned CV admissions were counted. It seems likely that unplanned admissions will have a greater effect on mortality than planned admissions, but this should be explored since the SCAF and AFFIRM studies could not adequately distinguish between them. If AF hospitalization, planned or unplanned, does not add much to the association of CV hospitalization with mortality, AF might even be excluded from the endpoint when aiming for a mortality substitute.
Although CV hospitalization is associated with increased risk of subsequent death, there are other reasons to consider it an important CV outcome. Atrial fibrillation can cause disabling symptoms requiring immediate treatment, but also other non-fatal CV events requiring hospitalization, such as stroke and (worsening of) heart failure. Non-fatal CV hospitalizations are associated with a decreased health-related quality of life in AF patients, which persisted for up to 2 years following admission.1 Furthermore, preventing CV hospitalizations will likely reduce healthcare costs since the majority of costs related to AF are attributable to inpatient expenditures, of which the first day of admission is the major inpatient cost driver.2 Shortening hospitalizations should add extra cost savings, and might lower mortality according to the SCAF analysis. Therefore, interventions that prevent or shorten CV hospitalizations, including AF-related, can be valuable to improve patients’ quality of life, reduce healthcare costs, and lower mortality. In the light of increasing trends in AF-related hospitalizations, reducing CV hospitalization is an important goal of therapy and would be an appropriate outcome for major clinical treatment trials in AF.9
Cardiovascular hospitalization seems a promising outcome for AF research, but when and how should it be used? An endpoint that occurs as frequently as CV hospitalization can be valuable in phase 2 trials, but can also be a rightful primary endpoint in phase 3 studies. The ATHENA study combined unplanned CV hospitalization with all-cause mortality for the primary endpoint, to account for competing risk, i.e. patients who are dead cannot be admitted to hospital. However, the limitations of CV hospitalization should be well understood for application of it in the correct settings. For example, from a practical perspective it needs to be considered how easily we can distinguish CV from non-CV hospitalization in a clinical trial. This needs to be carefully defined in the study protocol and blinded independent adjudication of CV and selected non-CV hospitalizations is desirable. Further, CV hospitalization depends on local practice and therefore varies among geographical regions and clinical settings, especially when including planned admissions. In a patient-randomized double-blind trial, this should not be a problem, but in cluster randomization trials with highly variable CV hospitalization rates among clinics and regions adjustment for clustering is essential.
In conclusion, CV hospitalization is a very reasonable primary outcome based on its impact on patients’ quality of life, costs of care, and mortality, and Friberg and Rosenqvist provide useful evidence for the latter. Ultimately, a sensitive and valid definition of CV hospitalization, based on combined evidence from multiple studies and varying clinical settings, will foster clinical AF research.
Conflict of interest: none declared.