Evaluation of Amiodarone Use for New-Onset Atrial Fibrillation in Critically Ill Patients With Septic Shock

Abstract

Objective: To describe the use of amiodarone in critically ill, septic shock patients experiencing new-onset atrial fibrillation (NOAF) during the acute resuscitative phase of septic shock. Methods: Single-center, retrospective review of adult medical or surgical intensive care unit (ICU) patients with septic shock and NOAF. All patients received amiodarone for NOAF during the acute resuscitative phase of septic shock. The cohort was analyzed via descriptive statistics. Associations between amiodarone exposure and clinical outcomes were analyzed via a Cox proportional-hazards model. An a priori defined sensitivity analysis of hospital survivors was also employed. Main Results: A total of 239 patients were included in the analysis. Patients had a median baseline Charlson Comorbidity Index of 4 (interquartile range [IQR]: 2-6) and were acutely ill with a median Acute Physiology and Chronic Health Evaluation II (APACHE II) score of 18 (IQR: 13-22) and an incidence of mechanical ventilation of 85%. In-hospital mortality was 56% with median ICU and hospital length of stay (LOS) of 9 and 15 days, respectively. Included patients received a median of 2760 (IQR: 1110-6415) mg of intravenous (IV) amiodarone during their ICU stay. Receipt of more than or equal to 2700 mg of amiodarone was identified as an independent factor associated with longer ICU LOS (hazard ratio [HR]: 1.30; 95% confidence interval [CI], 1.10-2.28). In a sensitivity analysis of hospital survivors (n = 105), receipt of more than or equal to 2700 mg of amiodarone remained independently associated with longer ICU LOS (HR: 1.64; 95% CI, 1.05-2.58). Conclusions: Exposure to more than or equal to 2700 mg of amiodarone in the setting of NOAF and septic shock is positively correlated with longer ICU LOS. Identifying opportunities to limit amiodarone exposure and address/resolve potential precipitating causes of NOAF in this clinical scenario may reduce the morbidity associated with septic shock.

Background

Atrial fibrillation (AF) is among the most prevalent arrhythmias encountered in the intensive care unit (ICU).^1,2 Critically ill patients with no history of arrhythmias are at risk for the development of new-onset arrhythmias, such as new-onset atrial fibrillation (NOAF), due to many established risk factors, including septic shock.³ The development of NOAF has been associated with morbidity- and mortality-related outcomes including higher hospital mortality and longer ICU and hospital length of stay (LOS) in the critically ill.^1,2,4-9

Despite the poor clinical outcomes associated with NOAF in septic shock, there is a paucity of data regarding the most optimal treatment. Medications commonly used to manage AF like calcium channel blockers (CCBs) and beta blockers (BB) may be avoided in critically ill patients due to their negative effects on blood pressure and inotropy. Amiodarone, however, provides an alternative due to its mechanism as both a rhythm and rate control agent with fewer effects on hemodynamics and inotropy.^10,11 Prior studies suggest that in septic shock patients experiencing NOAF, amiodarone is among the most frequently used agents.^6,12 However, data describing its use, appropriateness, or effects on clinical outcomes in this population remain lacking and/or conflicting.^12,13 These reasons, coupled with amiodarone’s well-described risks for arrhythmias and hepatic, thyroid, pulmonary, and other organ toxicities that may contribute to patient outcomes, underline the need for further investigation into the details of amiodarone use in this population.

This study aimed to describe the use of amiodarone in critically ill, septic shock patients experiencing NOAF during the acute resuscitative phase of septic shock. Furthermore, the association between total ICU amiodarone exposure and ICU LOS in this patient population was investigated.

Methods

Study Design

A retrospective cohort study was conducted in adult septic shock patients who received intravenous (IV) or enteral (per os [PO]) amiodarone for the management of NOAF at a large, academic tertiary institution from September 2009 to July 2016. The study was approved by the local institutional review board.

Study Subjects

Patients were identified using a pharmacy informatics database query. Patients were eligible for inclusion if admitted to a medical or surgical ICU during the study period, above or equal to 18 years of age, had International Classification of Diseases (ICD) discharge codes consistent with septic shock and AF, and received IV or PO amiodarone for NOAF during the acute resuscitative phase of septic shock. Patients admitted to an ICU other than a medical or surgical ICU were excluded to avoid inherent differences in patient populations and practice patterns at our institution, as well as to mitigate the potential inclusion patients with mixed shock. Additionally, patients were excluded for any ICD code consistent with cardiac arrest or non-AF arrhythmia during their admission or for a previous history of cardiac arrhythmia(s). A retrospective review of patients’ ICD codes during any hospitalization within 6 months before admission facilitated identification of patients with a history of cardiac arrhythmias. Only the first episode that met all inclusion criteria was included.

Definitions and Covariates

All definitions were determined prospectively. New-onset AF was defined as treatment for AF in a patient without a prior history of cardiac arrhythmias. The presence of septic shock was screened via utilization of the ICD codes. Study investigators reviewed and came to a consensus on included ICD codes before screening patients for inclusion (Supplemental material).

New-onset AF was defined as occurring during the acute resuscitative phase of septic shock if more than or equal to 1 vasoactive agent(s) was required to maintain hemodynamic stability within 24 hours before or after amiodarone initiation. Patients’ medical records were reviewed individually to determine the most likely time of septic shock diagnosis. This time was determined via identification of common, initial interventions for septic shock, including administration of IV fluids, vasoactive agents, and/or antibiotics, as outlined in published guidelines.¹⁴ The fact that NOAF may have occurred before these interventions and/or as a result of progression to septic shock served as the rationale for beginning the acute phase of septic shock 24 hours before the time of septic shock diagnosis. The acute phase of septic shock ended at the first occurrence of death, discharge, or vasoactive discontinuation with a subsequent vasoactive-free period of 24 hours. Vasoactive agents included norepinephrine, epinephrine, dopamine, phenylephrine, dobutamine, milrinone, and vasopressin.

Total amiodarone exposure was defined as the cumulative amount of IV and PO amiodarone received during ICU admission in milligrams (mg) from acute onset of septic shock to patient expiration or discharge, whichever occurred first. Of note, the ordering of amiodarone continuous infusions at our institution is accompanied by administration instructions to initiate infusions at a rate of 1 mg/min for the first 6 hours, followed by a rate reduction to 0.5 mg/min for the following 18 hours. No standard protocols exist to guide subsequent titration, discontinuation, or transition from IV to PO therapy.

Culture-positive patients had positive bacterial culture results within 72 hours of the acute onset of septic shock. Appropriate empiric antibiotics were defined as the receipt of an anti-infective active in vitro against culprit bacteria recovered within 24 hours of eventual positive culture being drawn. All in vitro testing was performed in the institution’s microbiology laboratory and interpreted by microbiology technicians.

The primary objective was to describe the use of amiodarone in critically ill, septic shock patients experiencing NOAF during the acute resuscitative phase of septic shock. The secondary objective was to explore associations between ICU amiodarone exposure and ICU LOS in this population. To elucidate this relationship, it was determined a priori that a threshold near the median total ICU amiodarone exposure would be implemented. This was considered exploratory in nature as no data have previously explored the relationship in this setting.

Electronic medical records were reviewed via an informatics database query and manual chart review. Demographic data, Acute Physiology and Chronic Health Evaluation II (APACHE II) score, and hemodynamic and laboratory values were collected.^15,16 Other collected data included, but were not limited to, receipt of other rate/rhythm-control agents, receipt of appropriate antibiotics, and hospital/ICU LOS. Data related to amiodarone administration included the most recent serum potassium and magnesium levels, number of IV amiodarone boluses, individual amounts of PO and IV amiodarone, and the total amount of amiodarone received during ICU stay.

Statistical Analysis

Descriptive analyses were used to describe patient demographics and other clinical variables. The association of amiodarone exposure and other covariates on ICU LOS was performed via a Cox proportional-hazards model. Factors controlled for in the Cox model were determined a priori as being clinically relevant and/or plausible to affect ICU LOS. The covariates were tested for collinearity, and none were collinear using a variance inflation factors threshold of more than 10. The proportional-hazards assumption was tested using Schoenfeld residual tests. To account for survival bias, an a priori sensitivity analysis was performed in hospital survivors. Kaplan-Meier survival curves were used to examine the time to ICU discharge as a function of total amiodarone exposure more than or equal to 2700 mg and compared via log-rank test. Tests were 2-tailed and unpaired, and a P value less than .05 represented statistical significance. All statistical analyses were performed using SPSS, version 22 software (SPSS Inc., Chicago, Illinois).

Results

A total of 8278 patients were screened for inclusion, with 239 patients eligible for analysis after applying all exclusion criteria (Figure 1). Baseline characteristics of the cohort are summarized in Table 1. Both medical and surgical ICU patients were well represented. The entire cohort displayed moderate severity of illness and comorbidity scores, with 85% requiring mechanical ventilation. The population also exhibited a median ICU LOS of 9 days, hospital LOS of 15 days, and incidence of mortality of 56%.

Figure 1.

Patient flow diagram.

Note. ICD = International Classification of Diseases; AF = atrial fibrillation; SICU = surgical intensive care unit; MICU = medical intensive care unit; NICU = neurology/neurosurgery intensive care unit; CTICU = cardiothoracic intensive care unit; CICU = cardiac intensive care unit.

Table 1.

Baseline Characteristics of Critically Ill, Septic Shock Patients With New-Onset Atrial Fibrillation.

	All patients (n = 239)
Age, y	67 (59-74)
Male, No. (%)	142 (59)
Race, No. (%)
White	191 (80)
Black	38 (16)
Other	10 (4)
MICU admission, No. (%)	122 (51)
SICU admission, No. (%)	117 (49)
Height, cm	173 (164-180)
Weight, kg	90 (71-109)
BMI, kg/m2	30 (24-36)
APACHE II score	18 (13-22)
CCI	4 (2-6)
Mechanical ventilation, No. (%)	202 (85)
Hemodialysis, No. (%)	65 (27)
Serum potassiuma, mmol/L	4.1 (3.7-4.6)
Serum magnesiuma, mg/dL	2.0 (1.6-2.3)
Culture positive, No. (%)	84 (35)
Site(s) of infectionb, No. (%)
Blood	32/84 (38)
Intra-abdominal	9/84 (12)
Lung	23/84 (27)
Tissue/wound	5/84 (6)
Urine	22/84 (26)
Other	18/84 (21)
Appropriate empiric antibiotics, No. (%)	70/84 (83)
Intravenous fluids receivedc, mL	4730 (3175-7500)
Exposure to vasopressin, No. (%)	122 (51)
Acute vasopressin exposured,e, h	26 (11-48)
Exposure to dopamine, No. (%)	5 (2)
Acute dopamine exposured,e, h	39 (26-51)
Exposure to epinephrine, No. (%)	51 (21)
Acute epinephrine exposured,e, h	17 (4-46)
Exposure to norepinephrine, No. (%)	238 (99)
Acute norepinephrine exposured,e, h	53 (23-103)
Exposure to phenylephrine, No. (%)	58 (24)
Acute phenylephrine exposured,e, h	4.5 (2-14)
Exposure to milrinone, No. (%)	4 (2)
Acute milrinone exposurec,d, h	81 (42-121)
Exposure to dobutaminec,d, No. (%)	45 (19)
Acute dobutamine exposurec,d, h	26 (9-71)
Exposure to other AF therapiesf, No. (%)	158 (66)
Exposure only before amiodarone initiation, No. (%)	42 (26)
Exposure only after amiodarone initiation, No. (%)	83 (53)
Exposure before and after amiodarone initiation, No. (%)	33 (21)
Direct current cardioversion attempted, No. (%)	10 (4)
In-hospital mortality, No. (%)	134 (56)
Hospital LOS, d	15 (8-26)
ICU LOS, d	9 (5-18)

Note. All data reported as median (interquartile range [IQR]) unless otherwise noted. MICU = medical intensive care unit; SICU = surgical intensive care unit; BMI = body mass index; APACHE = Acute Physiology and Chronic Health Evaluation; CCI = Charlson Comorbidity Index; AF = atrial fibrillation; LOS = length of stay.

^aAll values within 24 hours of amiodarone initiation. Value nearest amiodarone initiation included in cases of multiple eligible values.

^bColumn totals may not correlate as patients may have had more than 1 site of infection.

^cRefers to fluid received in initial 24 hours of cohort entry.

^dAcute refers to the acute resuscitative phase of septic shock.

^eReflects acute exposure of only patients exposed to that vasoactive.

^fIncludes exposure to beta blockers, calcium channel blockers, and/or digoxin.

In the 84 (35%) patients with positive microbiological results, our results indicate blood (38%), lung (27%), or urine (26%) were the primary sources of septic shock. In those with positive-culture data, 70 (84%) received appropriate empiric antibiotic therapy. The median amount of IV fluids administered to included patients in the acute setting was 4730 milliliters (mLs). Finally, 238 (99%) patients were exposed to norepinephrine during the acute phase of septic shock. Other vasopressors were less represented in this population; however, exposure to vasopressin occurred in 122 (51%) of patients.

Sixty-six percent of patients were treated with other therapies for AF, which included BBs, CCBs, and/or digoxin. Twenty-six percent of patients were treated with other therapies only before amiodarone administration. Conversely, the majority (53%) of patients were treated with other therapies only after amiodarone administration. Finally, 21% of patients were treated with other therapies both before and after exposure to amiodarone. The median time from last receipt of an alternative AF medication and cohort entry was 62 hours, and 10 (4%) patients underwent direct current cardioversion.

Amiodarone-related characteristics are summarized in Table 2. No patients in the cohort received initial therapy with PO amiodarone. Two-hundred and twenty-four (94%) patients received an initial IV bolus of 150 mg amiodarone. The median number of boluses received was 2 during the ICU stay. The median total ICU IV amiodarone exposure amongst all patients was 2760 (interquartile range [IQR]: 1110-6415) mg. Of the 118 patients surviving to ICU discharge, 34 (29%) received amiodarone on the day of ICU discharge. Of those 34 patients, 22 (65%) received amiodarone on the day of hospital discharge. Thirteen patients died after ICU discharge and an additional 8 patients were restarted on amiodarone after ICU discharge. In total, of 105 patients surviving to hospital discharge, 30 (29%) received amiodarone on the day of hospital discharge.

Table 2.

Amiodarone Characteristics.

	All patients (n = 239)
Amiodarone boluses received in ICU	2 (1-4)
Amiodarone bolus dose, mg	150 (150-150)
Initial amiodarone IV infusion rate, mg/min	1 (1-1)
Total ICU IV amiodarone exposure, mg	2760 (1110-6415)
Total ICU enteral amiodarone exposure, mg	0 (0-3600)
Total ICU amiodarone exposure, mg
0-999 mg	47 (20)
1000-2999 mg	78 (33)
3000-7999 mg	66 (27)
≥8000 mg	48 (20)
Receipt of amiodarone on day of ICU discharge, No. (%)a	34 (29)
Receipt of amiodarone on day of hospital discharge, No. (%)b	30 (29)

Note. All data reported as median (interquartile range [IQR]), unless otherwise noted. ICU = intensive care unit; IV = intravenous.

^aData reported in relation to number of patients surviving to ICU discharge (n = 118).

^bData reported in relation to number of patients surviving to hospital discharge (n = 105).

Figure 2 describes the likelihood of remaining in the ICU following the onset of septic shock, as a function of total amiodarone exposure more than or equal to 2700 mg. Patients treated with more than or equal to 2700 mg of amiodarone had a median increase of 7 days in ICU LOS (12 days vs 5 days, P ≤ .01). When multiple covariates were adjusted for using a Cox model (Table 3), receipt of more than or equal to 2700 mg of amiodarone remained correlated with an increased likelihood of remaining in the ICU, with an adjusted hazard ratio of 1.30 (95% confidence interval [CI], 1.10-2.28). The need for mechanical ventilation, vasopressor exposure, APACHE II score, and lung infection also correlated with remaining in the ICU.

Figure 2.

Impact of total amiodarone exposure on intensive care unit length of stay.

Note. ICU = intensive care unit; LOS = length of stay.

Table 3.

Independent Factors Associated With Intensive Care Unit Length of Stay.

	Hazard ratio	95% confidence interval	P value
Appropriate empiric antibiotics	0.49	0.31-0.78	<.01
APACHE II score	1.05	1.01-1.08	<.01
Hemodialysis	1.39	0.86-2.22	.18
Total amiodarone ≥2700 mg	1.30	1.10-2.28	<.01
Lung infection	3.21	1.52-6.81	<.01
Mechanical ventilation	4.00	2.33-6.71	<.01
Total vasopressor exposure, h	1.03	1.01-1.04	<.01
Acute IV fluids received, mL	0.99	0.94-1.05	.81

Note. Variables were selected for entry into the Cox proportional-hazards model, a priori, if they were felt to be biologically linked and/or clinically relevant to ICU length of stay. APACHE = Acute Physiology and Chronic Health Evaluation; ICU = intensive care unit; IV = intravenous.

Hazard ratios greater than 1 indicate that the factor is associated with a greater probability of remaining in the ICU. Hazard ratios of less than 1 indicate variables that correlate with a greater chance of being discharged from the ICU sooner.

In a sensitivity analysis including only septic shock hospital survivors (n = 105), the adjusted hazard ratio associated with receipt of more than or equal to 2700 mg of total amiodarone exposure equaled 1.64 (95% CI, 1.05-2.58; P < .01). Adjusted hazard ratios related to the need for mechanical ventilation (hazard ratio [HR]: 3.28; 95% CI, 1.84-5.84; P < .01) and lung infection (HR: 4.05; 95% CI, 1.73-9.49; P = .01) exhibited associations with longer ICU LOS, whereas receipt of appropriate empiric antibiotics remained independently linked to shorter ICU LOS (HR: 0.55; 95% CI, 0.34-0.90; P = .02) in this population.

Discussion

This retrospective analysis of septic shock patients who developed NOAF and were treated with amiodarone represents an acutely ill, commonly seen population from which conclusions and hypotheses can be drawn. The cohort’s acuity is accentuated via the high incidence of mechanical ventilation, requirement of hemodynamic support, more than 30% need for renal replacement therapy, and elevated Charlson Comorbidity Index (CCI) and APACHE II scores. The reported mortality rate of 56% underscores this point and is higher than common mortality predictive tools would have estimated. While this study was not designed to assess differences between those with and without NOAF in the setting of septic shock, it is worth noting that previous studies have demonstrated greater hospital mortality and longer ICU LOS in those patients who develop NOAF than those who do not, and logistic regression has revealed NOAF invokes a 1.6 times higher likelihood of hospital mortality after controlling for numerous AF risk factors and confounders.^4,6,13

Patients in the studied population displayed baseline and treatment characteristics to be expected for septic shock. Approximately 30% incidence of positive cultures, most often of blood, urine, and/or lung source(s) are representative of a septic shock population. The most commonly used vasopressors in this population, norepinephrine and vasopressin, are consistent with current guideline recommendations for the treatment of septic shock unresponsive to IV fluid resuscitation.¹⁴

The treatment patterns of NOAF in the setting of sepsis and/or septic shock in the presented study are consistent with previous evidence that suggests multiple interventions are often used in this clinical scenario. Meierhenrich et al described a septic shock population that most often received combination therapy for NOAF; however, Meierhenrich et al⁶ found that amiodarone was the most frequently used medication amongst the 49 patients experiencing NOAF (73%). In contrast, Shaver et al¹³ suggested that among their 123 NOAF patients (85% diagnosed with shock), treatment with BBs (47%) or diltiazem (47%) was the most common, followed by amiodarone (40%). Finally, in the aforementioned billing data analysis, investigators concluded that amiodarone was the most likely to be used among patients who were critically ill with septic shock, as well as in NOAF, likely due to the favorable impact on hemodynamics.¹² Given the heterogeneity in study design and patient populations of previous studies, results are difficult to apply and generalize.

The details provided about amiodarone use in this population, to the best knowledge of the investigators, have not been previously described. The results suggest most patients received initial IV treatment, but did not receive a complete, 10-g load of amiodarone. Given the lack of data surrounding the optimal approach to both short- and long-term treatment of NOAF in this setting, it is worthwhile to note that nearly 30% of patients received amiodarone on their final day of ICU and hospital admission. The clinical implications of amiodarone use in this setting remain to be elucidated. Ongoing assessment and treatment of precipitating causes of NOAF, along with review and discussion of amiodarone continuation is warranted. These steps are likely to reduce the incidence of patients being discharged on inappropriate therapy which could predispose them to adverse effects associated with amiodarone use.

This retrospective analysis demonstrates that exposure to more than or equal to 2700 mg amiodarone was correlated with a 1.30 times higher likelihood of remaining in the ICU. This effect was independent of many established confounders. To our knowledge, no prior study has attempted to discern the association between amiodarone exposure and ICU LOS in the setting of septic shock and NOAF. In a previously published, propensity-matched study of 5388 patients with a shock diagnosis who received a vasoactive medication on the same day as an AF medication, patients who received amiodarone experienced higher hospital mortality than patients receiving BBs (42 vs 27%, P < .001).¹¹ A subsequent sensitivity analysis validated this increased risk in the setting of NOAF (relative risk [RR]: 0.67; 95% CI, 0.59-0.77). Furthermore, a small, digoxin-controlled study suggested digoxin as the first-line therapy for recent-onset AF or atrial flutter due to its better safety profile than amiodarone.¹⁷ The presented results are unique compared to these analyses by focusing on the impact of a common intervention for these patients, rather than the disease state itself and using an endpoint other than mortality. These points, along with amiodarone’s well-described organ toxicities, drug interactions, potential to cause arrhythmias, and the increasing cost of a day in the ICU, speak to the importance of understanding whether a relationship between amiodarone use and clinical outcomes exist.

Identifying the potential for confounders and biases a priori is what led investigators to design this trial around a Cox proportional-hazards model and subsequent sensitivity analysis of only hospital survivors. The sensitivity analysis found the likelihood of remaining in the ICU remained significant in relation to patients who received more than or equal to 2700 mg of amiodarone, suggesting a 1.64 times higher likelihood of remaining in the ICU. Taken in this context, our data support the notion that septic shock patients who develop NOAF may benefit from judicious use of amiodarone in this setting. While the investigators advise careful interpretation of these data, practitioners may consider limiting the exposure to amiodarone in this setting as clinically indicated.

Despite our results, this study has several important limitations to consider. First, it remains unclear whether the association between increased amiodarone exposure and longer ICU LOS is due to the increased exposure to amiodarone being a marker of greater severity of illness. Furthermore, as we used an easily identifiable time point of ICU LOS as the cutoff for cumulative amiodarone exposure, ICU LOS and cumulative amiodarone exposure may be inherently related. This, however, was not evident in our statistical assessment for collinearity. Future trials may consider alternative cutoffs for cumulative amiodarone exposure such as patient expiration, hospital discharge, or end of septic shock. While this association does not imply causality, this is still the first study to our knowledge that attempts to address optimal management in this cohort of critically ill, septic shock patients who develop NOAF and the importance of managing the underlying causes of NOAF. Related to this, the necessity to treat NOAF in septic shock patients remains pertinent and unaddressed by the presented study. To minimize bias, we used previous publications and determined definitions prospectively. We also relied on ICD codes for inclusion and exclusion, as well as for important study definitions. In doing so, we may have failed to identify and/or inappropriately identified patients with septic shock and/or NOAF. We are unable to determine the clinical benefit and/or futility of amiodarone use related to restoration of normal sinus rhythm, as these data are not readily and reliably available in our electronic medical record. This study does not account for well-described differences in the pharmacokinetics and pharmacodynamics of IV and PO amiodarone. At present, little consensus exists, particularly in the setting of NOAF related to septic shock in how these differences effect outcomes. Another limitation, which is common to all single-center studies, is the external validity of the findings. Finally, intricate data related to vasoactive use, IV fluid administration, volume status, and other therapeutic interventions are lacking in the presented model. The lack of inclusion of such data is, largely, a downfall of our retrospective design. However, given our inclusion criteria related to vasoactive use, and our institution’s guideline-based protocol for the treatment of sepsis, we feel that these confounders were likely consistent across groups.¹⁴ The investigators recognize that any statistical model of this nature will have limitations, but believe that we identified major factors that could have driven outcomes.

Our results focus on a commonly used agent for NOAF in the setting of septic shock and its effect on morbidity. Given the intuitive appeal of amiodarone in this scenario, it is important to understand all potential sequelae of this common intervention. These findings may serve as a reference and/or starting point for future investigations, as much is yet to be discovered about common treatment strategies for NOAF in the septic shock population.

Conclusions

In a cohort of critically ill adults experiencing NOAF and septic shock, a median of 2 IV boluses and 2760 mg of total IV amiodarone was used during their ICU stay. Exposure to more than or equal to 2700 mg of amiodarone was positively correlated with longer ICU LOS after adjustment for confounders and in a sensitivity analysis of hospital survivors. Further assessment of and consideration of the need for treatment with amiodarone in septic shock patients who develop NOAF is warranted.