Impact of Rhythm-control Therapy for New-onset Atrial Fibrillation in Critically ill Patients : A Post Hoc Analysis From the Prospective Observational AFTER-ICU study

Sustained new-onset atrial brillation (AF) is time-dependently associated with hospital mortality. However, whether rhythm-control therapy can achieve sinus rhythm (SR) restoration in critically ill patients is unknown. This study aimed to assess the impact of rhythm-control therapy on SR restoration in critically ill patients with new-onset AF. This study is a post hoc analysis of the AFTER-ICU study, a prospective observational study of patients with new-onset AF in 32 Japan intensive care units. This study included patients with and without rhythm-control therapy with new-onset AF. A multivariable analysis was performed using Cox proportional hazards regression analysis including rhythm-control therapy as a time-varying covariate for SR restoration. CHADS2, point: heart failure, hypertension, years, diabetes mellitus; 2 points: transient ischemic attack or a prior ACE, angiotensin converting enzyme; ARBs, angiotensin II receptor blockers; APACHE II, Acute Physiology and Chronic Health Evaluation II; SOFA, Sequential Organ Failure Assessment; MV, mechanical ventilation; RRT, renal replacement therapy.


Variables and measurement
To assess the impact of rhythm-control therapy on SR restoration, we compared patients with rhythmcontrol therapy to those without. We obtained the following information from the AFTER-ICU study: patient demographics, physiological data, and drugs used at AF onset. We also obtained the following information within 7 days after initial AF onset or during ICU stay, whichever was shorter: timing of directcurrent cardioversion, drugs used for new-onset AF, adverse events (bleeding events or cardiac arrhythmia other than AF), and timing of cardiac rhythm transition. The rhythm-control drugs for new-onset AF were magnesium sulfate, amiodarone, pilsicainide, aprindine, cibenzoline, adenosine triphosphate, disopyramide, ecainide, bepridil, and lidocaine. The rate control drugs were beta-blocking agents (landiolol, bisoprolol, propranolol, and carvedilol), calcium-channel blockers (diltiazem and verapamil), and digoxin. We also de ned the use of rhythm-control drugs and/or undergoing direct-cardioversion as rhythm-control therapy.

Outcomes
Our primary outcome was the last SR restoration within 7 days after the initial AF onset or during the ICU stay, whichever was shorter. SR restoration was de ned as sustained SR for longer than 24 hours after the conversion from AF to SR. If the patients were discharged from ICU with SR within 24 hours after the conversion of cardiac rhythm from AF to SR, they were also de ned as those with SR restoration. The secondary outcomes were the patients' cardiac rhythm at ICU discharge, AF duration, ICU length of stay, hospital length of stay, adverse events, ICU mortality, hospital mortality, and in-hospital stroke. In-hospital stroke was de ned as symptomatic cerebral infarction diagnosed by a neurologist or a neurosurgeon or determined via new computed tomography or magnetic resonance imaging ndings [14]. The de nition of the other collected variables is detailed in Additional table 1.

Statistical analysis
The study results are presented as median and interquartile range or as absolute numbers with percentage, as appropriate. In all analyses, the number of missing data was reported, and cases with missing data were excluded from each analysis. Comparisons between the two groups were conducted using chi-square test or Fisher's exact test for categorical variables and Mann-Whitney U test for continuous variables. P<0.05 was considered statistically significant.
To assess the association between rhythm-control therapy and SR restoration, we modeled the time from AF onset to SR restoration using Cox proportional hazards regression. Patients who were later initiated on rhythm-control therapy might have longer time until SR restoration. To address time-related bias, we used the rhythm-control therapy as a time-varying covariate in this model. The following variables were included in this model according to their clinical relevance and importance in previous studies [3][4][5]8,9,11,18,30,31]: age, previous history of congestive heart failure, patient category (nonscheduled surgical, scheduled surgical, and medical), Acute Physiology and Chronic Health Evaluation (APACHE) II scores [32] at ICU admission, infection at AF onset, renal replacement therapy at AF onset, mechanical ventilation at AF onset, administration of drugs (any vasopressors, inotropes, and dexmedetomidine) at AF onset, heart rate at AF onset, and the laboratory data (potassium and white blood cells) before AF onset. To account for the nonlinear effects of age, heart rate at AF onset, potassium, and white blood cells on outcomes, the penalized smoothing spline function was incorporated into the Cox proportional hazards model. Patients who were discharged from the ICU or died with remaining AF within 7 days after AF onset were censored because we could not measure their duration until SR restoration. The Cox proportional hazards regression analyses were performed using R version 3.5.1 (The R Foundation for Statistical Computing, Vienna, Austria). All other analyses were performed using Stata version 16 (StataCorp, College Station, TX, USA).

Sensitivity analyses
Although direct-current cardioversion is a rhythm-control therapy, undergoing DC direct-current cardioversion does not have the effect of sustained SR, which is different from the pharmacologic interventions. Therefore, we performed sensitivity analysis using Cox proportional hazards regression for the rhythm-control therapy without direct-current cardioversion.

Results
A total of 423 patients with new-onset AF were enrolled in the AFTER-ICU study, of whom 178 (42%) were treated with the rhythm-control therapy for new-onset AF. The initial timing of rhythm-control therapy during 7 days after AF onset is shown in Fig. 1. Among patients with the rhythm-control therapy, 151 (36%) underwent rhythm-control therapy within 24 hours after AF onset and 131 (31%) within 6 hours after AF onset.
The patients' demographic and clinical characteristics are shown in Table 1. Laboratory data are shown in Additional table 2. Almost two-thirds of the study patients were medical patients and had infection at AF onset. Patients in the rhythm-control group were younger than those in the non-rhythm-control group.
The APACHE II score and the proportion of patients who required mechanical ventilation at AF onset were greater in the rhythm-control group than those in the non-rhythm-control group. The proportion of those with previous histories of congestive heart failure and ischemic heart disease did not statistically differ between the two groups.
The physiological data before and at AF onset are shown in Table 2. At AF onset, although there was no difference in the mean arterial pressure at AF onset between two groups, the rhythm-control group had higher heart rate and used vasopressors and inotropes more frequently than the non-rhythm-control group.
Interventions for new-onset AF and outcomes are shown in Table 3. The combinations of the interventions for each patient in the rhythm-control group are shown in Additional g. 1. Among the patients who had undergone direct-current cardioversion for new-onset AF, 37 (57%) received rhythmcontrol drugs. There were a few patients who only used a single drug, and magnesium sulfate was the most frequently used rhythm-control drug. Regarding the rate control drugs, beta-blocking agents were used more frequently in the rhythm-control group than in the non-rhythm-control group. Although the rhythm-control group had longer AF duration than the non-rhythm-control group, the proportion of patients who remained in AF at ICU discharge in the rhythm-control group was signi cantly lower than that in the non-rhythm-control group. Meanwhile, the length of hospital stay, hospital mortality, and frequency of adverse events in the rhythm-control group were higher than those in the non-rhythm-control group.
The results of the Cox models for SR restoration after adjustment for the prespeci ed confounding factors are shown in Table 4 and Additional g. 2. Among all covariates included in this model, only rhythm-control therapy had a signi cant positive association with SR restoration.
The sensitivity analyses are shown in Additional table 3 and g.3. The results of these analyses were similar to those of the main analysis.

Key ndings
This post hoc analysis of a prospective multicenter observational study assessed the impact of rhythmcontrol therapy on SR restoration in critically ill patients. Among 423 patients with new-onset AF, 178 (42%) were treated with rhythm-control therapy. Patients in the rhythm-control group had higher APACHE II score and more frequently required mechanical ventilation, vasopressors, and inotropes at AF onset than those in the non-rhythm-control group. ICU and hospital mortality in the rhythm-control group were also greater than those in the non-rhythm-control group. Among the rhythm-control drugs, many of which were administered within 6 hours after AF onset, magnesium sulfate was the most frequently used. Although the total AF duration in the rhythm-control group was longer than that in the non-rhythm-control group, the patients in the rhythm-control group remained in AF at ICU discharge less frequently than those in the non-rhythm-control group. Moreover, the multivariable Cox regression analysis that included the rhythm-control therapy as a time-varying covariate con rmed the association between rhythm-control therapy and SR restoration.

Relationship with previous studies
For an interventional study to assess whether rhythm-control therapy can improve patient outcomes, we need the premise that rhythm-control therapy is effective for SR restoration. In critically ill patients, the e cacy of magnesium sulfate and amiodarone has been evaluated for new-onset AF in a few interventional studies conducted in the 2000s [15,16,21,33]. For example, a single-center RCT with 60 patients with AF compared the e cacy of amiodarone with that of diltiazem [33]. They reported that amiodarone tended to be more effective than diltiazem on SR restoration at 4 hours after AF onset (42.5% vs. 30%, P=0.34). Another single-center RCT with 42 patients with AF without hemodynamic instability reported that sulfate magnesium was better than amiodarone for SR restoration at 24 hours after AF onset (78% vs. 50%) [21]. Moreover, a prospective single-arm study reported that the combination of amiodarone and magnesium sulfate might have a high probability of SR restoration (90% at 24 hours after AF onset) [22]. Consistent with these previous studies, we found that rhythm-control therapy, including magnesium sulfate and amiodarone, was associated with SR restoration. Our current analyses took into account the duration until the last SR restoration from all AFs (including recurrent AF) within 7 days after AF onset. Moreover, among variables included in the multivariable regression analysis, only rhythm-control therapy showed a signi cant positive impact on SR restoration. Therefore, our study highlighted the importance of rhythm-control therapy for SR restoration from new-onset AF in critically ill patients.
Previous observational studies for rhythm-control therapy had methodological problems for evaluation of rhythm-control therapy, especially for observation time points of cardiac rhythms (mainly at 24 hours) [15][16][17][18]. Such speci c observation time points cannot consider the time-varying nature of rhythm-control therapy, often called "immortal bias" [34][35][36]. Because rhythm-control therapy is generally initiated for sustained AF, patients with rhythm-control therapy may have longer AF duration and lower chance of SR restoration at a speci c time point than those without rhythm control. In fact, in our study, although the rhythm-control group had a longer AF duration in the univariable analysis, the multivariable analysis, which used rhythm-control therapy as a time-varying covariate, showed the signi cant impact of the therapy on SR restoration. This statistical approach is less common in critical care research [34,36]. These ndings suggest the importance of appropriately treating time-varying covariates for conducting observational studies in critical care setting.

Significance and implications
Despite the lack of evidence, rhythm-control therapy for AF has been generally indicated in hemodynamically unstable patients [17,18,37]. Therefore, the rhythm-control group in observational studies might have poor outcomes [38]. Indeed, we found that the rhythm-control group had greater ICU and hospital mortality with greater severity scores and higher proportion of patients requiring mechanical ventilation and vasopressors at AF onset. To avoid this confounder, RCTs for the assessment of rhythmcontrol therapy are warranted. In addition, we also presented the timing for the initiation of rhythm-control therapy, rhythm-control drug options, proportion of patients who remained in AF, and frequency of adverse events. These ndings may provide important insight for the design and feasibility of interventional studies assessing rhythm-control therapy in new-onset AF.

Strengths and limitations
To the best of our knowledge, the current study included a larger number of critically ill patients with newonset AF than that of previous studies that assessed the impact of rhythm-control therapy [4,5,15,16]. With the multivariable regression analysis considering time-related bias, we showed a signi cant impact of rhythm-control therapy on SR restoration. However, this study also has several limitations. First, because it was conducted only in Japan, our ndings may have limited generalizability. However, the rate of SR restoration at ICU discharge in the rhythm-control group was 85%, which was within the range found in previous studies [15,29]. Second, we could not determine the best timing of rhythm control for SR restoration. In our study, almost 70% of all rhythm-control therapies were initiated within 6 hours after AF onset, which may be the appropriate duration for initiating the treatment in future interventional studies. Third, most patients with rhythmcontrol therapy also received rate control therapy, which might have contributed to SR restoration. Future studies that include a speci c protocol for drug usage are needed. Fourth, we did not distinguish directcurrent cardioversion from other pharmacologic interventions as rhythm-control therapy because whether there is any difference in impact on SR restoration between those interventions is unknown. However, our sensitivity analyses without direct-current cardioversion showed similar results as the primary ndings. An analysis focused on direct-current cardioversion using our database seems warranted. Finally, we could not identify a speci c intervention in favor of SR restoration because of the various rhythm-control drugs used. Sulfate magnesium, which appears to have a low risk of adverse events [15,18,39], was the most frequently used among the rhythm-control drugs. This drug may be promising for future interventional studies assessing rhythm-control therapy.

Conclusion
This study showed that rhythm-control therapy was associated with SR restoration for new-onset AF in critically ill patients. Because patients treated with rhythm control in observational studies may have poor outcomes due to selection bias (i.e., patients with hemodynamic instability tend to get the treatment), further interventional studies for rhythm-control therapy are strongly warranted to avoid this confounder.

Declarations
Ethics approval and consent to participate: The AFTER-ICU study was registered at UMIN-CTR (UMIN000026401), and the original study protocol was approved by the Jikei University Institutional Review Board (approval no. 28-200[8443]) and the ethics committees of all other participating hospitals, with an opt-out policy from the patient or proxy.
Consent for publication: Not applicable.
Availability of data and materials: The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.
Competing interests: The authors declare that they have no competing interests.