You are on page 1of 9

European Journal of Internal Medicine xxx (xxxx) xxx–xxx

Contents lists available at ScienceDirect

European Journal of Internal Medicine


journal homepage: www.elsevier.com/locate/ejim

Predicting spontaneous conversion to sinus rhythm in symptomatic atrial


fibrillation: The ReSinus score
Jan Niederdöckl1, Alexander Simon2, Filippo Cacioppo1, Nina Buchtele3,4, Anne Merrelaar1,
Nikola Schütz1, Sebastian Schnaubelt1, Alexander O Spiel1,2, Dominik Roth1,
Christian Schörgenhofer3, Harald Herkner1, , Hans Domanovits1, Michael Schwameis1

1
Department of Emergency Medicine, Medical University of Vienna, Währinger Gürtel 18-20, 1090 Vienna, Austria
2
Zentrale Notaufnahme, Wilhelminenspital, Montleartstr.37, 1160 Vienna, Austria
3
Department of Clinical Pharmacology, Medical University of Vienna, Währinger Gürtel 18-20, 1090 Vienna, Austria
4
Department of Medicine I, Medical University of Vienna, Währinger Gürtel 18-20, 1090 Vienna, Austria

ARTICLE INFO ABSTRACT

Keywords: Background: The optimal management of patients presenting to the Emergency Department with hemodyna-
symptomatic atrial fibrillation mically stable symptomatic atrial fibrillation remains unclear. We aimed to develop and validate an easy-to-use
spontaneous conversion score to predict the individual probability of spontaneous conversion to sinus rhythm in these patients
prediction score Methods: This retrospective cohort study analyzed 2426 cases of first-detected or recurrent hemodynamically
derivation
stable non-permanent symptomatic atrial fibrillation documented between January 2011 and January 2019 in
validation
an Austrian academic Emergency Department atrial fibrillation registry. Multivariable analysis was used to
develop and validate a prediction score for spontaneous conversion to sinus rhythm during Emergency
Department visit. Clinical usefulness of the score was assessed using decision curve analysis
Results: 1420 cases were included in the derivation cohort (68years, 57-76; 43% female), 1006 cases were
included in the validation cohort (69years, 58-76; 47% female). Six variables independently predicted sponta-
neous conversion. These included: duration of atrial fibrillation symptoms (<24hours), lack of prior cardio-
version history, heart rate at admission (>125bpm), potassium replacement at K+ level ≤3.9mmol/l, NT-
proBNP (<1300pg/ml) and lactate dehydrogenase level (<200U/l). A risk score weight was assigned to each
variable allowing classification into low (0-2), medium (3-5) and moderate (6-8) probability of spontaneous
conversion. The final score showed good calibration (p=0.44 and 0.40) and discrimination in both cohorts (c-
indices: 0.74 and 0.67) and clinical net benefit
Conclusion: The ReSinus score, which predicts spontaneous conversion to sinus rhythm, was developed and
validated in a large cohort of patients with hemodynamically stable non-permanent symptomatic atrial fi-
brillation and showed good calibration, discrimination and usefulness
Registration: NCT03272620

1. Introduction Conversely, patients who are likely to convert to sinus rhythm spon-
taneously, but undergo early pharmacological or electrical cardiover-
In patients with symptomatic atrial fibrillation rates of spontaneous sion may be unnecessarily hospitalized and exposed to the risk of post-
conversion are high.1-3 In real-world emergency cohorts optimal man- conversion arrhythmias and complications by general anesthesia or
agement is unclear; however, treatment with early cardioversion is a antiarrhythmic drugs.7, 8
common practice, although the individual probability of a patient Given the rising prevalence9 and economic burden10 of atrial fi-
converting to sinus rhythm spontaneously is uncertain. Deferred car- brillation, several decision aids have been developed to facilitate the
dioversion in patients with a low probability of spontaneous conversion estimation of a patients’ individual stroke, bleeding and mortality risk,
may result in an unjustified treatment delay and increased risk of thus guiding safe and effective long-term management.11, 12 A tool to
stroke, heart failure and progression to persistent atrial fibrillation,4-6 estimate the individual probability of spontaneous conversion of


Corresponding author. Tel.: +4314040019640, Fax: +4314040019650.
E-mail address: harald.herkner@meduniwien.ac.at (H. Herkner).

https://doi.org/10.1016/j.ejim.2020.07.022
Received 20 February 2020; Received in revised form 14 July 2020; Accepted 31 July 2020
0953-6205/ © 2020 The Author(s). Published by Elsevier B.V. on behalf of European Federation of Internal Medicine. This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/BY-NC-ND/4.0/).

Please cite this article as: Jan Niederdöckl, et al., European Journal of Internal Medicine, https://doi.org/10.1016/j.ejim.2020.07.022
J. Niederdöckl, et al. European Journal of Internal Medicine xxx (xxxx) xxx–xxx

patients presenting with symptomatic atrial fibrillation, however, is not planned using a random sample for model derivation and one for va-
available. Ideally, informed decisions of patients and clinicians should lidation in a 1:1 sample ratio. However, owing to the considerations of
be based on this information. Steyerberg and Verguowe16 on the benefits of temporal validation we
In this study we developed and validated an easy-to-use score to restricted the validation sample to the most contemporary cases at a
estimate the individual probability of spontaneous conversion to sinus sample size sufficient to handle up to 10 independent predictors in the
rhythm in adult patients with hemodynamically stable non-permanent model. Accordingly, we aimed at a sample size of approximately 1000
symptomatic atrial fibrillation. patients for the validation set. Temporal validation assesses the per-
formance of a model by applying it to the most recent data from the
2. Material and methods population it was initially derived. Thus, it refers to the generalizability
and transportability of study findings to plausibly related populations
2.1. Study design/setting and is considered a more robust test for prediction models. A compar-
ison of the set of excluded patients (n=345) with the set of patients
This retrospective cohort study is based on an atrial fibrillation finally included in the validation cohort (n=1006) is available with the
registry including consecutive adult patients with atrial fibrillation supplement (Supp. Table 4).
treated at the Emergency Department at the Medical University of Spontaneous conversion was defined as return to sinus rhythm
Vienna, an academic tertiary care facility. The Emergency Department during Emergency Department visit without any attempt at pharma-
comprises an outpatient care section and an affiliated critical care unit cological or electrical cardioversion. Rate-control therapy as well as
covering more than 90 000 patients overall 13 and about 600 patients electrolyte replacement was not rated as an attempt at cardioversion.
with atrial fibrillation per year. Patients presenting with atrial fi- Patients who converted spontaneously did not receive any anti-ar-
brillation and hypokalemia (less than 3.5mmol/l) or low-normal po- rhythmic drugs prior to restoration of sinus rhythm.
tassium levels (3.5-3.9mmol/l) receive potassium replacement with
Elozell Spezial intravenous solution (containing 24 mmol potassium 2.4. Statistical methods
and 12mmol magnesium per 250ml) based on provider-dependent
practice. Acute rate and rhythm control therapy is based on current Variables are presented as absolute values (n), relative frequencies
European Society of Cardiology (ESC) guidelines 14 and depends on the (%) and median (25-75% interquartile ranges, IQR). Between-group
patient's left ventricular function, hemodynamics, drug history in- comparisons were performed using the Mann-Whitney U test for con-
cluding anticoagulation status and comorbidities. Rate control medi- tinuous variables or the chi-squared test/Fisher's exact test for nominal
cation includes beta-adrenergic receptor blockers (metoprolol, es- variables. Univariable logistic regression with spontaneous conversion
molol), verapamil and cardiac gylcosides (digoxin, digitoxin). to sinus rhythm as a dependent variable was performed on available
Pharmacological cardioversion is performed mainly with vernakalant, cases. Several candidate predictors for spontaneous conversion, which
ibutilide and amiodarone. Standard biphasic defibrillators are used for were judged to be clinically plausible, were tested. Only variables
synchronized direct current electrical cardioversion. available within one hour upon admission were used to allow the
The study was approved by the Ethics Committee of the Medical creation of a risk score readily useable in clinical practice. Continuous
University of Vienna and conducted in accordance with ICH-GCP variables in univariable analysis were categorized by selecting clinically
guidelines and Helsinki Declaration. Patients or the public were not relevant cut-offs, which were the closest to the statistically optimal cut-
involved in the design, or conduct, or reporting, or dissemination of our offs, and examined for linear and non-linear associations including re-
research. The study is registered with clinicaltrials.gov stricted cubic splines. Categorization of variables yielded at parsimony
(NCT03272620). and dichotomy, and cut-offs were optimized for maximum discrimina-
tion in the derivation cohort.
2.2. Atrial fibrillation registry Variables significant in univariable analysis were entered in a
multivariable logistic regression model to calculate adjusted odds ratio
The atrial fibrillation registry has previously been described in de- (OR) with 95% confidence intervals (95% CI). A stepwise process was
tail.15 In brief, the registry started in January 2011 and prospectively used, aiming at the most parsimonious model. Interaction was assessed
includes all cases of atrial fibrillation confirmed by 12-lead electro- using the likelihood ratio test. In case of interaction the interaction
cardiography. Prior to inclusion informed consent is obtained. Demo- terms were used in subsequent models. Multicollinearity between the
graphics, past medical history including comorbidities, concomitant variables in the final model was assessed as ill-conditioning, i.e. the
medication and previous attempts at electrical cardioversion, the impact of small random changes (perturbations) to variables on para-
CHA2DS2VASC score, results from blood gas analysis, blood count, meter estimates. Firth regression penalizing the log-likelihood with
chemistry, coagulation variables, thyroid function, troponin and NT- one-half of the logarithm of the determinant of the information matrix
proBNP levels are obtained. Additionally, vital signs including heart was used for the final model to avoid overfitting to the derivation data.
rate, blood pressure and oxygen saturation, symptoms attributable to These adjusted coefficients of significant multivariable predictors were
atrial fibrillation, the time of symptom onset, the type of atrial fi- then divided by the lowest coefficient value in the model and rounded
brillation, and treatments including electrolyte substitution, rate con- to the nearest integer. From these values a risk score weight was as-
trol medication and cardioversion attempts are documented by study signed to each predictor in the model. The sum of risk scores for each
fellows. patient was subsequently calculated.
Calibration was assessed by performing the Hosmer–Lemeshow
2.3. The ReSinus score goodness-of-fit test and by plotting observed vs. predicted incidence
rate across categories of the developed score. In the derivation set, as
The score was developed and validated in cases of hemodynamically well as in the validation set, the predictive performance of the risk score
stable first-detected or recurrent non-permanent symptomatic atrial was assessed using the c-statistic, which was 1000-fold cross-validated
fibrillation entered into the atrial fibrillation registry between January by bootstrapping in order to evaluate the discriminative ability.
2011 and January 2019 (Figure 1). To test the robustness of the model according to the time to cardi-
AF, atrial fibrillation. oversion we performed a sensitivity analysis restricting the observation
Patients with a history of permanent atrial fibrillation and atrial time from admission to restoration of sinus rhythm to two hours.
fibrillation events occurring in the context of critical illness were ex- Clinical usefulness was evaluated using decision curve analysis.17 In
cluded before cohort allocation. Initially, internal validation was this context, clinical net benefit is the relationship between the benefit

2
J. Niederdöckl, et al. European Journal of Internal Medicine xxx (xxxx) xxx–xxx

Figure 1. Study flow chart. Out of 3012 cases of symptomatic atrial fibrillation documented between 2011 and 2019, a total of 2426 cases of hemodynamically stable
first-detected or recurrent non-permanent symptomatic atrial fibrillation were eligible for analysis. Owing to the benefits of temporal validation the validation sample
was restricted to the most contemporary cases at a sample size sufficient to handle up to 10 independent predictors in the model. 1420 cases were finally included in
the derivation cohort and 1006 cases were included in the validation cohort.

of treating those, who need treatment, and the harm of treating those, baseline characteristics and clinical course of both study cohorts.
who do not need treatment. Decision curve analysis allows the eva- 968 patients (67.3%) in the derivation cohort and 717 patients
luation of clinical net benefit of a prognostic tool over a range of (71.3%) in the validation cohort who did not convert spontaneously
threshold probabilities of having a positive outcome. Clinical net ben- underwent an attempt of pharmacological/electrical cardioversion.
The median duration from admission to restoration of sinus rhythm
true positives false positives p
efit is calculated as n n
( 1 tp ) , where n is the total
was significantly longer in patients who were cardioverted than in those
t
number of patients, and ptis the threshold probability of having a po-
sitive outcome. True and false positives are calculated using pt as the who converted spontaneously to sinus rhythm: 4.1 vs. 2.1 hours in the
cut-point for determining a positive or negative result. This calculation derivation cohort (p<0.001) and 3.7 vs. 1.5 hours in the validation
is repeated over a range of clinically meaningful threshold prob- cohort (p=0.040) (Supp. Table 5).
abilities.18
The analyses followed the framework of derivation and validation of 3.1. Derivation of the ReSinus score
prediction models proposed by Steyerberg and Vergouwe.16 Reporting
followed the TRIPOD statement.19 Missing data were included as se- Spontaneous conversion to sinus rhythm occurred in 186 patients
parate categories for each variable as appropriate. Stata 14 (StataCorp, (13%). Clinical and laboratory characteristics of patients in the deri-
College Station, TX, USA) was used for data analysis. Generally, a two- vation cohort stratified by the occurrence of spontaneous conversion
sided p-value <0.05 was considered statistically significant. are shown in Table 2.
Univariable analysis identified six variables independently asso-
3. Results ciated with spontaneous conversion to sinus rhythm. The identified
values of predictors were doubled and rounded to the nearest integer
Out of 3012 cases of symptomatic atrial fibrillation entered into the (Table 3).
registry between 2011 and 2019, 2426 cases of hemodynamically stable The weighted score included the duration of atrial fibrillation re-
first-detected or recurrent non-permanent symptomatic atrial fibrilla- lated symptoms (<24 hours; 2 points), a lack of prior cardioversion
tion were eligible for analysis. 1420 cases were included in the deri- history (2 points), heart rate at admission (>125 beats per minute; 1
vation cohort (median age 68 years, IQR 57 - 76; 43% female) and 1006 point), potassium replacement at K+ level ≤3.9mmol/l (1 point), NT-
cases were included in the validation cohort (median age 69 years, IQR proBNP (<1300 pg/ml; 1 point) and lactate dehydrogenase level
58 - 76; 47% female). Table 1 provides an overview of demographics, (<200 U/l; 1 point), totaled to a maximum score of 8 points. The factor

3
J. Niederdöckl, et al. European Journal of Internal Medicine xxx (xxxx) xxx–xxx

Table 1
Demographics and baseline characteristics of the derivation and the validation cohort
Derivation cohort Validation cohort p=
n=1420 n=1006
available* (n) available* (n)

Clinical Characteristics
Age, years (IQR) 68 (57 - 76) 1420 69 (58 - 76) 1420 0.190
Female gender, n (%) 615 (43) 472 (47) 0.065
BMI, kg/m˄2 (IQR) 28 (25 - 30) 1420 28 (24 - 30) 984 0.824
BP systolic, mmHG (IQR) 134 (120 - 150) 1290 134 (120 - 150) 967 0.876
BP diastolic, mmHG (IQR) 82 (72 - 95) 1279 85 (75 - 95) 962 0.034
Heart rate, bpm (IQR) 130 (111 - 147) 845 130 (111 - 146) 625 0.762
Comorbidities
Lone AF, n (%) 86 (6) 72 (7) 0.311
Heart failure, n (%) 242 (17) 174 (17) 0.894
Hypertension, n (%) 873 (61) 593 (59) 0.185
DM, n (%) 327 (23) 130 (13) 0.433
TIA, n (%) 38 (3) 12 (1) 0.016
Stroke, n (%) 93 (7) 49 (5) 0.061
CAD, n (%) 243 (17) 166 (17) 0.711
Previous myocardial infarcton, n (%) 110 (8) 91 (9) 0.276
PAD, n (%) 59 (4) 39 (4) 0.733
COPD, n (%) 114 (8) 72 (7) 0.452
Valvular pathology, n (%) 349 (25) 235 (23) 0.445
Hyperlipidaemia, n (%) 452 (32) 301 (30) 0.331
Hyperthyreosis, n (%) 56 (4) 46 (5) 0.580
Hypothyreosis, n (%) 221 (16) 163 (16) 0.759
Current smoker, n (%) 77 (5) 44 (4) 0.749
AF history
First AF episode, n (%) 217 (15) 183 (18) 0.505
Previous AF episodes, n (%) 3 (0 - 8) 691 3 (0 - 8) 506 0.373
Previous electrical CV, n (%) 398 (28) 298 (30) 0.392
Ablation, n (%) 229 (16) 180 (18) 0.227
Duration of AF symptoms, h (IQR) 6 (2 - 21) 760 6 (2 - 19) 558 0.997
CHA2DS2-VASc (IQR) 3 (1 - 4) 1339 2 (1 - 4) 961 0.029
Medication
Phenprocoumon, n (%) 355 (25) 198 (20) 0.002
DOACS, n (%) 245 (17) 213 (21) 0.015
Cardiac glycosides, n (%) 56 (4) 38 (4) 0.834
Beta blocker, n (%) 625 (44) 437 (43) 0.779
Amiodarone, n (%) 255 (18) 147 (15) 0.039
AT-2 Blocker, n (%) 316 (22) 266 (16) 0.025
ACE Blocker, n (%) 284 (20) 174 (17) 0.102
Pre-Treatment
Potassium replacement, n (%) 712 (50) 377 (37) 0.171
Laboratory
Haematocrit, % (IQR) 41 (38 - 45) 1351 42 (38 - 45) 941 0.134
WBC, G/l (IQR) 8 (7 - 10) 1363 8 (7 - 10) 950 0.860
Creatinine, mg/dl (IQR) 1 (1 - 1) 1361 1 (1 - 1) 954 0.350
Potassium, mmol/l (IQR) 4.0 (3.8 - 4.3) 1321 4.0 (3.8 - 4.3) 930 0.973
LDH, U/l (IQR) 196 (167 - 238) 1117 198 (169 - 240) 792 0.435
NT-proBNP, pg/ml (IQR) 943 (303 - 2393) 1172 981 (312 - 2739) 831 0.355
hs-Troponin T, ng/l (IQR) 14 (8 - 25) 1090 14 (8 - 27) 821 0.663
CRP, mg/dl (IQR) 0 (0 - 1) 1318 0 (0 - 1) 927 0.750
Lactate, mmol/l (IQR) 1 (1 - 2) 1238 2 (1 - 2) 870 0.226
INR, (IQR) 2 (1 - 3) 738 1 (1 - 2) 527 <0.001
TSH, μU/ml (IQR) 2 (1 - 3) 980 2 (1 - 3) 691 0.492

*Number of patients who had the variable available. Missing data for some score variables in both cohorts ranged from 0 to 21%.
ACE (angiotensin converting enzyme), AF (atrial fibrillation), AT (angiotensin), BMI (body mass index), BP (blood pressure), CAD (coronary artery disease), COPD
(chronic obstructive pulmonary disease), CRP (C-reactive protein), CV (cardioversion), DM (diabetes mellitus), DOAC (direct oral anticoagulants), hs (high-sensitive),
INR (international normalized ratio), LA (left atrium), LDH (lactate dehydrogenase), NT-proBNP (N-terminal-pro brain natriuretic peptide), PAD (peripheral artery
disease), RA (right atrium), TIA (transient ischemic attack), TSH (thyroid stimulating hormone), WBC (white blood cells).
The frequency distribution of individual DOACS, glycosides and beta-blockers is available with the supplement (Supp. Table 1).

‘potassium replacement at K+ level ≤3.9mmol/l’ included all patients values showed good calibration (p=0.43) (Figure 2).
who received potassium substitution unless potassium level exceeded The cumulative final score value of each patient allowed classifi-
3.9mmol/l. The observed incidence of short-term spontaneous conver- cation into low (0-2), medium (3-5) and moderate (6-8) probability of
sion was 50% for patients with 8 points and 0% for patients with 0 spontaneous conversion. The rates of short-term spontaneous conver-
points. The odds ratio associated with every increase of one score point sion to sinus rhythm across these three classes were 3.8%, 12.3%,
was 1.61 (95% CI 1.47–1.76; p < 0.001). 33.8% respectively (Figure 3).
The final model showed good discrimination (c-index 0.74; 95% CI
0.70–0.77). The p-value of the Hosmer-Lemeshow goodness-of-fit test
3.2. Validation of the ReSinus score
was 0.43. The predicted and observed incidence of short-term sponta-
neous conversion to sinus rhythm across the groups built by final score
The observed incidence of spontaneous conversion in the validation

4
J. Niederdöckl, et al. European Journal of Internal Medicine xxx (xxxx) xxx–xxx

Table 2
Clinical and biomarker characteristics among patients with and without spontaneous conversion
Spontaneous No Spontaneous p=
Conversion n=186 Conversion n=1234
available* (n) available* (n)

Clinical Characteristics
Age, years (IQR) 68 (56 - 76) 186 68 (58 - 75) 1234 0.490
Male gender, n (%) 108 (58) 697 (56) 0.626
BMI, kg/m˄2 (IQR) 28 (25 - 30) 186 28 (2 - 30) 1234 0.315
BP systolic, mmHG (IQR) 133 (120 - 150) 168 134 (11 - 150) 1122 0.728
BP diastolic, mmHG (IQR) 82 (72 - 95) 167 82 (7 - 95) 1112 0.825
Heart rate, bpm (IQR) 136 (117 - 149) 142 129 (110 - 146) 703 0.012
Comorbidities
Lone AF, n (%) 9 (5) 77 (6) 0.405
Heart failure, n (%) 19 (10) 223 (18) 0.008
Hypertension, n (%) 128 (69) 754 (61) 0.022
DM, n (%) 20 (11) 177 (14) 0.217
TIA, n (%) 5 (3) 33 (3) 0.964
Stroke, n (%) 12 (6) 81 (7) 0.996
CAD, n (%) 33 (18) 210 (17) 0.769
Previous myocardial 14 (8) 96 (8) 0.896
infarction, n (%)
PAD, n (%) 4 (2) 55 (4) 0.137
COPD, n (%) 12 (6) 102 (8) 0.329
Valvular pathology, n (%) 46 (25) 303 (25) 0.999
Hyperlipidaemia, n (%) 67 (36) 385 (31) 0.166
Hyperthyreosis, n (%) 6 (3) 50 (4) 0.674
Hypothyreosis, n (%) 37 (20) 184 (15) 0.083
Current smoker, n (%) 22 (12) 135 (11) 0.868
AF history
First AF episode, n (%) 40 (22) 177 (14) 0.027
Previous AF episodes, n 2 (0 - 5) 97 3 (0 - 8) 594 0.058
(IQR)
Previous electrical CV, n 38 (20) 360 (29) 0.018
(IQR)
Ablation, n (%) 31 (17) 198 (16) 0.802
Duration of AF symptoms, h 4 (2 - 8) 133 7 (2 - 24) 627 <0.001
(IQR)
CHA2DS2-VASc (IQR) 3 (2 - 4) 182 2 (1 - 4) 1157 0.366
Medication
Phenprocoumon, n (%) 38 (20) 317 (26) 0.140
DOACS, n (%) 35 (19) 210 (17) 0.545
Cardiac glycosides, n (%) 6 (3) 50 (4) 0.589
Beta blocker, n (%) 92 (49) 533 (43) 0.108
Amiodarone, n (%) 22 (12) 233 (19) 0.022
AT-2 Blocker, n (%) 54 (29) 262 (21) 0.018
ACE Blocker, n (%) 46 (25) 238 (19) <0.001
Pre-Treatment
Potassium replacement, n 124 (67) 588 (48) <0.001
(%)
Laboratory
Haematocrit, % (IQR) 41 (37 - 44) 180 42 (38 - 45) 1171 0.237
WBC, G/l (IQR) 8 (7 - 10) 183 8 (7 - 10) 1180 0.950
Creatinine, mg/dl (IQR) 1 (1 - 1) 183 1 (1 - 1) 1178 0.094
Potassium, mmol/l (IQR) 3.9 (3.7 - 4.1) 181 4.0 (3.8 - 4.3) 1140 <0.001
LDH, U/l (IQR) 190 (163 - 220) 173 198 (167 - 241) 944 0.012
NT-proBNP, pg/ml (IQR) 516 (22 - 1282) 164 1042 (334 - 2574) 1008 <0.001
hs-Troponin T, ng/l (IQR) 15 (8 - 25) 143 14 (8 - 25) 947 0.406
CRP, mg/dl (IQR) 0 (0 - 1) 176 0 (0 - 1) 1142 0.752
Lactate, mmol/l (IQR) 1 (1 - 2) 177 1 (1 - 2) 1061 0.374
INR, (IQR) 1 (1 - 2) 97 2 (1 - 3) 641 <0.001
TSH, μU/ml (IQR) 2 (1- 3) 136 2 (1 - 3) 844 0.456

*Number of patients who had the variable available. Missing data for some score variables in both cohorts ranged from 0 to 21%.
ACE (angiotensin converting enzyme), AF (atrial fibrillation), AT (angiotensin), BMI (body mass index), BP (blood pressure), CAD (coronary artery disease), COPD
(chronic obstructive pulmonary disease), CRP (C-reactive protein), CV (cardioversion), DM (diabetes mellitus), DOAC (direct oral anticoagulants), hs (high-sensitive),
INR (international normalized ratio), LA (left atrium), LDH (lactate dehydrogenase), NT-proBNP (N-terminal-pro brain natriuretic peptide), PAD (peripheral artery
disease), RA (right atrium), TIA (transient ischemic attack), TSH (thyroid stimulating hormone), WBC (white blood cells).
The frequency distribution of individual DOACS, glycosides and beta-blockers is available with the supplement (Supp. Table 2).

cohort according to their classification (low, medium, moderate) was benefit of using the ReSinus score over the full spectrum of possible
7.9%, 9.1% and 19.6%. The score showed good calibration (p=0.40) thresholds (Figure 4).
and discrimination (c statistic 0.67; 95% CI 0.63–0.72) (Figure 2). Figure 5 provides an easy-to-use template for pragmatic calculation
The sensitivity analysis censoring the observation time at two hours of the ReSinus score in the clinical setting. Online calculation of the
suggested robustness of the model (Supp. Table 3). score is available at www.meduniwien.ac.at/notfall/resinus.
The decision curve analysis indicated a substantial clinical net

5
J. Niederdöckl, et al. European Journal of Internal Medicine xxx (xxxx) xxx–xxx

Table 3
Independent predictors of spontaneous conversion to sinus rhythm
Predictor Coefficient 95% CI p= Weighted Score

Duration of AF symptoms <24 h 0.89 ( 0.53 - 1.25 ) <0.001 2


No previous electrical CV 0.75 ( 0.35 - 1.15 ) <0.001 2
Heart rate > 125 bpm 0.55 ( 0.22 - 0.88 ) <0.001 1
Potassium replacement at K+ level ≤3.9mmol/l 0.55 ( 0.21 - 0.89 ) 0.002 1
NT-proBNP < 1300 pg/ml 0.54 ( 0.19 - 0.88 ) 0.003 1
LDH < 200 U/l 0.45 ( 0.11 - 0.78 ) 0.009 1

Firth regression was used to penalize the log-likelihood of the model to avoid overfitting to the derivation data. AF (atrial fibrillation), CV (cardioversion), NT-
proBNP (N-terminal-pro brain natriuretic peptide), LDH (lactate dehydrogenase).

Figure 2. Observed and predicted incidence of spontaneous conversion in (A) the derivation cohort and (B) the validation cohort. Calibration was visualized by
plotting observed vs. predicted incidence rate across categories of the developed score in the derivation set (A), as well as in the validation set (B). The dotted line
represents perfect calibration, the solid line represents actual calibration.

4. Discussion within one hour of admission were used. Six independent predictors of
spontaneous conversion to sinus rhythm were identified including four
In this study we developed and validated an easy-to-use risk score to clinical (actual heart rate, duration of atrial fibrillation related symp-
estimate the individual probability of spontaneous conversion to sinus toms, clinical history of previous cardioversion, potassium replace-
rhythm in patients with hemodynamically stable non-permanent ment) and two routine laboratory variables (NT-proBNP and LDH
symptomatic atrial fibrillation. The score includes both clinical and level). The final score was well calibrated, showed good discriminative
laboratory information and was developed from a cohort of patients ability and clinical usefulness.
treated in a real-world emergency setting. To allow for the creation of a
risk score readily applicable in clinical practice, only variables available

Figure 3. Observed incidence of spontaneous conversion across low (0-2), medium (3-5) and moderate (6-8) probability categories for spontaneous conversion in the
(A) derivation cohort and (B) the validation cohort.

6
J. Niederdöckl, et al. European Journal of Internal Medicine xxx (xxxx) xxx–xxx

Figure 4. Decision curve analysis showing clinical usefulness of


the ReSinus score. X-axis depicts threshold probability of sponta-
neous conversion. Y-axis depicts clinical net benefit of three dif-
ferent strategies: Dashed line: ReSinus score. Solid line: assume all
patients will convert. Thin line: assume no patient will convert.
ReSinus score has a positive net benefit over the whole spectrum
of threshold probabilities. This was especially true in the clinically
relevant probability range from 5 to 30%, which corresponds to
the spontaneous conversion rates observed in our and many other
studies.

Figure 5. Stratification according to the probability of spontaneous conversion using the ReSinus score. Work along criteria A-F from the middle to the edge. Each
corresponding answer leads to the adjacent field of the next circle (blue = true; white = false). The final score can be read directly from the outermost circle. The
individual probability of spontaneous conversion according to the score is given by the bar on the right side. AF (atrial fibrillation), CV (cardioversion), NT-proBNP
(N-terminal-pro brain natriuretic peptide), LDH (lactate dehydrogenase).

4.1. Clinical predictors score weight alongside a lacking history of cardioversion. Several
variables of the ReSinus score reflect advanced heart disease and may
Atrial fibrillation is considered a continuous process of electro- indicate structural atrial alterations. In this context, a history of pre-
anatomical remodeling promoting electrical dissociation, heterogenic vious cardioversion attempts may suggest the presence of structural
conduction and arrhythmic atrial contractions.20, 21 With progressive changes, which impede spontaneous conversion and prompt the need
disease, accumulating structural changes in atrial myocytes promote for cardioversion. Likewise, a low heart rate may indicate that the
further episodes of atrial fibrillation, which facilitates electro-physio- disease is already advanced. Data on the relation between heart rate
logic conditions that perpetuate atrial arrhythmia and limit the prob- and the probability of spontaneous conversion in atrial fibrillation are
ability of spontaneous conversion.22-24 It has been shown experimen- lacking. However, a higher heart rate has previously been shown to
tally that the duration of atrial fibrillation negatively affects its favor successful cardioversion and is associated with both a lower rate
spontaneous conversion probability.25 Accordingly, in the ReSinus of progression 26 and a lower rate of recurrence in atrial fibrillation. In
score the duration of atrial fibrillation was assigned the highest risk more advanced disease, remodeling processes within atria and the AV-

7
J. Niederdöckl, et al. European Journal of Internal Medicine xxx (xxxx) xxx–xxx

node, as well as negative chrono-and dromotropic treatment, may affiliated critical care unit allowing fast diagnostic work-up and the
promote lower ventricular conduction rates, which may thus be asso- opportunity to rapidly perform cardioversion at any time. As this was a
ciated with a lower probability of spontaneous conversion. registry-based study, time to cardioversion attempts and observation
periods were not standardized. It is thus conceivable that patients as-
4.2. Laboratory predictors signed to the non-spontaneous conversion cohort would have converted
spontaneously over the following hours or days if they had not under-
The final score includes two biomarkers. NT-pro BNP has previously gone early cardioversion to restore sinus rhythm. Some related bias
been shown to be associated with atrial appendage dysfunction27 and cannot be excluded and should be considered when interpreting our
the risk of both new-onset and recurrent atrial fibrillation.28-32 Atrial findings. However, sensitivity analysis restricting the observational
natriuretic peptide is specifically expressed in the atria and may thus be time-window from admission to two hours suggested robustness of the
an even more sensitive indicator of structural atrial disease33 but this is model. Furthermore, the median duration from admission to restoration
not yet routinely available in clinical practice. Lactate dehydrogenase of sinus rhythm was significantly longer in patients who were finally
in contrast is a widely used unspecific marker of tissue damage, which cardioverted than in those who converted spontaneously. At our in-
is elevated in various medical conditions. It has recently been shown stitution, cardioversion of hemodynamically stable symptomatic AF
that lactate cascades play a crucial role in the process of atrial re- patients, who do not early convert spontaneously, is performed as soon
modeling and in the maintenance of atrial fibrillation.34 We can only as exam results are available and a trial of fluid and/or potassium
speculate as to whether elevated LDH levels in our study patients reflect challenge has been performed to improve the probability of sponta-
atrial alterations associated with atrial fibrillation or other comorbid- neous conversion. This might be a common approach and reflect real-
ities. It is, however, evident that in atrial fibrillation biomarker-guided world practice in emergency settings, where ED-overcrowding and
management lags behind that of other major cardiovascular diseases limited space for observation are common conditions. Yet, a future
including myocardial infarction and heart failure, where both diagnosis prospective trial should include a predefined observation schedule and
and treatment rely on laboratory values. The inclusion of biomarkers follow-up period. In addition, the single-center design limits our results’
into prediction models for the assessment of bleeding, stroke and death generalizability to different settings and populations. We followed the
risk in patients with atrial fibrillation has substantially improved their methodological guidance by Steyerberg and Verguowe and used the
accuracy and usefulness.35-37 It is conceivable that the inclusion of most contemporaneous set for validating our score. Geographical or
further biomarkers not yet readily available in clinical practice would strong external validation may provide more robust model validation,
likewise improve the performance of the ReSinus score. but was not be performed in this single-center study. Further validation
of the developed score is needed to conclusively assess its performance.
4.3. Application of the ReSinus score As not only hypokalemia but also potassium levels in the low-
normal range may be considered target conditions for potassium re-
The ReSinus score was developed and validated in a real-world placement, we included the combined information of low/low-normal
cohort of patients with non-permanent symptomatic atrial fibrillation potassium level (≤3.9mmol/l) and potassium substitution in the
presenting to the Emergency Department. Atrial fibrillation is among model. Patients with potassium levels exceeding 3.9mmol/l were not
the most frequent arrhythmias encountered in the acute care sector and counted as potassium replacement. This needs to be noted, as clinical
places an increasingly critical economic burden on health care sys- practice on potassium substitution in atrial fibrillation may vary be-
tems,38, 39 mainly driven by treatment associated complications and tween institutions and clinicians.
costs of hospitalization.14 Emergency department visits and hospitali- Finally, possible bias arising from missing data cannot be fully ex-
zations for atrial fibrillation are increasing,10, 40 stressing the im- cluded. For some of the score variables data were missing in the deri-
portance of strategies aimed at personalizing acute management deci- vation and in the validation set. We avoided strong assumptions on the
sions to reduce hospital admissions for unnecessary cardioversions, missing data and included these as separate categories into our models.
treatment complications and associated health care costs. Yet, it needs to be acknowledged that this does not exclude bias asso-
2016 ESC guidelines suggest immediate cardioversion in patients ciated with data missing not at random. Our results thus should be
with recent-onset atrial fibrillation and related symptom burden or interpreted with appropriated caution.
cardiorespiratory compromise. In hemodynamically stable patients
with symptomatic atrial fibrillation, however, decisions as to whether 5. Conclusions
immediate restoration of sinus rhythm in the individual patient is ne-
cessary, or whether spontaneous conversion can be expected, remains The ReSinus score, which uses clinical and routine laboratory
challenging in clinical practice because no objective decision guidance variables to predict spontaneous conversion to sinus rhythm, was de-
is available. An easy-to-use decision tool may offer the opportunity to veloped and validated in a large cohort of patients with hemodynami-
personalize patient treatment already in a very early phase, facilitate cally stable non-permanent symptomatic atrial fibrillation, showed
shared decision-making processes, fasten acute therapeutic manage- good calibration, discrimination and clinical usefulness. Once vali-
ment and may reduce the length of stay in crowded Emergency dated, its implementation in clinical practice may aid physicians in
Departments. The ReSinus score may especially support decisions on guiding early cardioversion decisions while reducing the risk of over-
early cardioversion in patients in whom timely spontaneous conversion treatment.
to sinus rhythm is unlikely to occur. Conversely, in patients with high
probability of spontaneous conversion, the ReSinus score could Authors
strengthen delayed-cardioversion approaches as suggested by
Pluymaekers et al and help avoiding treatment associated complica- Jan Niederdoeckl, Alexander Simon, Filippo Cacioppo, Nina
tions.3 Buchtele, Anne Merrelaar, Nikola Schütz, Sebastian Schnaubelt,
Alexander O. Spiel, Dominik Roth, Christian Schörgenhofer, Harald
4.4. Strengths and Limitations Herkner, Hans Domanovits, Michael Schwameis

The particular strength of this study is its real-world-cohort struc- Declaration of interests
ture. This suggests robust results that could prove valid beyond highly
selected study populations. The Emergency Department at the Medical The authors declare that they have no known competing financial
University of Vienna comprises an outpatient care section and an interests or personal relationships that could have appeared to

8
J. Niederdöckl, et al. European Journal of Internal Medicine xxx (xxxx) xxx–xxx

influence the work reported in this paper. [19] Collins GS, Reitsma JB, Altman DG, Moons KG. Transparent Reporting of a multi-
The authors declare the following financial interests/personal re- variable prediction model for Individual Prognosis or Diagnosis (TRIPOD): the
TRIPOD statement. Annals of internal medicine 2015;162:55–63.
lationships which may be considered as potential competing interests: [20] Allessie MA, de Groot NM, Houben RP, Schotten U, Boersma E, Smeets JL, et al.
Electropathological substrate of long-standing persistent atrial fibrillation in pa-
Declarations of interest tients with structural heart disease: longitudinal dissociation. Circulation
Arrhythmia and electrophysiology 2010;3:606–15.
[21] Nattel S, Harada M. Atrial remodeling and atrial fibrillation: recent advances and
none translational perspectives. Journal of the American College of Cardiology
2014;63:2335–45.
[22] Ausma J, Wijffels M, Thone F, Wouters L, Allessie M, Borgers M. Structural changes
Supplementary materials of atrial myocardium due to sustained atrial fibrillation in the goat. Circulation
1997;96:3157–63.
Supplementary material associated with this article can be found, in [23] Thijssen VL, Ausma J, Liu GS, Allessie MA, van Eys GJ, Borgers M. Structural
changes of atrial myocardium during chronic atrial fibrillation. Cardiovascular
the online version, at doi:10.1016/j.ejim.2020.07.022.
pathology: the official journal of the Society for Cardiovascular Pathology
2000;9:17–28.
References [24] Zhang L, Huang B, Scherlag BJ, Ritchey JW, Embi AA, Hu J, et al. Structural
changes in the progression of atrial fibrillation: potential role of glycogen and fi-
brosis as perpetuating factors. International journal of clinical and experimental
[1] Cotter G, Blatt A, Kaluski E, Metzkor-Cotter E, Koren M, Litinski I, et al. Conversion pathology 2015;8:1712–8.
of recent onset paroxysmal atrial fibrillation to normal sinus rhythm: the effect of no [25] Wijffels MC, Kirchhof CJ, Dorland R, Allessie MA. Atrial fibrillation begets atrial
treatment and high-dose amiodarone. A randomized, placebo-controlled study. fibrillation. A study in awake chronically instrumented goats. Circulation
European heart journal 1999;20:1833–42. 1995;92:1954–68.
[2] Decker WW, Smars PA, Vaidyanathan L, Goyal DG, Boie ET, Stead LG, et al. A [26] Padfield GJ, Steinberg C, Swampillai J, Qian H, Connolly SJ, Dorian P, et al.
prospective, randomized trial of an emergency department observation unit for Progression of paroxysmal to persistent atrial fibrillation: 10-year follow-up in the
acute onset atrial fibrillation. Annals of emergency medicine 2008;52:322–8. Canadian Registry of Atrial Fibrillation. Heart rhythm 2017;14:801–7.
[3] Pluymaekers N, Dudink E, Luermans J, Meeder JG, Lenderink T, Widdershoven J, [27] Yu GI, Cho KI, Kim HS, Heo JH, Cha TJ. Association between the N-terminal plasma
et al. Early or Delayed Cardioversion in Recent-Onset Atrial Fibrillation. The New brain natriuretic peptide levels or elevated left ventricular filling pressure and
England journal of medicine 2019;380:1499–508. thromboembolic risk in patients with non-valvular atrial fibrillation. Journal of
[4] Nuotio I, Hartikainen JE, Gronberg T, Biancari F, Airaksinen KE. Time to cardio- cardiology 2016;68:110–6.
version for acute atrial fibrillation and thromboembolic complications. Jama [28] Takase H, Dohi Y, Sonoda H, Kimura G. Prediction of Atrial Fibrillation by B-type
2014;312:647–9. Natriuretic Peptide. Journal of atrial fibrillation 2013;5:674.
[5] Anter E, Jessup M, Callans DJ. Atrial fibrillation and heart failure: treatment con- [29] Sinner MF, Stepas KA, Moser CB, Krijthe BP, Aspelund T, Sotoodehnia N, et al. B-
siderations for a dual epidemic. Circulation 2009;119:2516–25. type natriuretic peptide and C-reactive protein in the prediction of atrial fibrillation
[6] Nieuwlaat R, Capucci A, Camm AJ, Olsson SB, Andresen D, Davies DW, et al. Atrial risk: the CHARGE-AF Consortium of community-based cohort studies. Europace:
fibrillation management: a prospective survey in ESC member countries: the Euro European pacing, arrhythmias, and cardiac electrophysiology: journal of the
Heart Survey on Atrial Fibrillation. European heart journal 2005;26:2422–34. working groups on cardiac pacing, arrhythmias, and cardiac cellular electro-
[7] Gallagher MM, Yap YG, Padula M, Ward DE, Rowland E, Camm AJ. Arrhythmic physiology of the European Society of Cardiology 2014;16:1426–33.
complications of electrical cardioversion: relationship to shock energy. [30] Smith JG, Newton-Cheh C, Almgren P, Struck J, Morgenthaler NG, Bergmann A,
International journal of cardiology 2008;123:307–12. et al. Assessment of conventional cardiovascular risk factors and multiple bio-
[8] Wyse DG, Waldo AL, DiMarco JP, Domanski MJ, Rosenberg Y, Schron EB, et al. A markers for the prediction of incident heart failure and atrial fibrillation. Journal of
comparison of rate control and rhythm control in patients with atrial fibrillation. the American College of Cardiology 2010;56:1712–9.
The New England journal of medicine 2002;347:1825–33. [31] Svennberg E, Lindahl B, Berglund L, Eggers KM, Venge P, Zethelius B, et al. NT-
[9] Krijthe BP, Kunst A, Benjamin EJ, Lip GY, Franco OH, Hofman A, et al. Projections proBNP is a powerful predictor for incident atrial fibrillation - Validation of a
on the number of individuals with atrial fibrillation in the European Union, from multimarker approach. International journal of cardiology 2016;223:74–81.
2000 to 2060. European heart journal 2013;34:2746–51. [32] Zografos TA, Katritsis DG. Natriuretic Peptides as Predictors of Atrial Fibrillation
[10] Rozen G, Hosseini SM, Kaadan MI, Biton Y, Heist EK, Vangel M, et al. Emergency Recurrences Following Electrical Cardioversion. Arrhythmia & electrophysiology
Department Visits for Atrial Fibrillation in the United States: Trends in Admission review 2013;2:109–14.
Rates and Economic Burden From 2007 to 2014. Journal of the American Heart [33] Buttner P, Schumacher K, Dinov B, Zeynalova S, Sommer P, Bollmann A, et al. Role
Association 2018;7. of NT-proANP and NT-proBNP in patients with atrial fibrillation: Association with
[11] Lane DA, Lip GY. Use of the CHA(2)DS(2)-VASc and HAS-BLED scores to aid de- atrial fibrillation progression phenotypes. Heart rhythm 2018;15:1132–7.
cision making for thromboprophylaxis in nonvalvular atrial fibrillation. Circulation [34] Xu J, Xu X, Si L, Xue L, Zhang S, Qin J, et al. Intracellular lactate signaling cascade
2012;126:860–5. in atrial remodeling of mitral valvular patients with atrial fibrillation. Journal of
[12] Lip GY, Nieuwlaat R, Pisters R, Lane DA, Crijns HJ. Refining clinical risk stratifi- cardiothoracic surgery 2013;8:34.
cation for predicting stroke and thromboembolism in atrial fibrillation using a novel [35] Hijazi Z, Lindback J, Alexander JH, Hanna M, Held C, Hylek EM, et al. The ABC
risk factor-based approach: the euro heart survey on atrial fibrillation. Chest (age, biomarkers, clinical history) stroke risk score: a biomarker-based risk score for
2010;137:263–72. predicting stroke in atrial fibrillation. European heart journal 2016;37:1582–90.
[13] Schwameis M, Buchtele N, Schober A, Schoergenhofer C, Quehenberger P, Jilma B. [36] Hijazi Z, Oldgren J, Lindback J, Alexander JH, Connolly SJ, Eikelboom JW, et al.
Prognosis of overt disseminated intravascular coagulation in patients admitted to a The novel biomarker-based ABC (age, biomarkers, clinical history)-bleeding risk
medical emergency department. European journal of emergency medicine: official score for patients with atrial fibrillation: a derivation and validation study. Lancet
journal of the European Society for Emergency Medicine 2017;24:340–6. 2016;387:2302–11.
[14] Kirchhof P, Benussi S, Kotecha D, Ahlsson A, Atar D, Casadei B, et al. 2016 ESC [37] Hijazi Z, Oldgren J, Lindback J, Alexander JH, Connolly SJ, Eikelboom JW, et al. A
Guidelines for the management of atrial fibrillation developed in collaboration with biomarker-based risk score to predict death in patients with atrial fibrillation: the
EACTS. European journal of cardio-thoracic surgery: official journal of the ABC (age, biomarkers, clinical history) death risk score. European heart journal
European Association for Cardio-thoracic Surgery 2016;50:e1–88. 2018;39:477–85.
[15] Niederdockl J, Simon A, Schnaubelt S, Schuetz N, Laggner R, Sulzgruber P, et al. [38] Lee WC, Lamas GA, Balu S, Spalding J, Wang Q, Pashos CL. Direct treatment cost of
Cardiac biomarkers predict mortality in emergency patients presenting with atrial atrial fibrillation in the elderly American population: a Medicare perspective.
fibrillation. Heart 2019;105:482–8. Journal of medical economics 2008;11:281–98.
[16] Steyerberg EW, Vergouwe Y. Towards better clinical prediction models: seven steps [39] Ringborg A, Nieuwlaat R, Lindgren P, Jonsson B, Fidan D, Maggioni AP, et al. Costs
for development and an ABCD for validation. European heart journal of atrial fibrillation in five European countries: results from the Euro Heart Survey
2014;35:1925–31. on atrial fibrillation. Europace: European pacing, arrhythmias, and cardiac elec-
[17] Steyerberg EW, Vickers AJ, Cook NR, Gerds T, Gonen M, Obuchowski N, et al. trophysiology: journal of the working groups on cardiac pacing, arrhythmias, and
Assessing the performance of prediction models: a framework for traditional and cardiac cellular electrophysiology of the European Society of Cardiology
novel measures. Epidemiology 2010;21:128–38. 2008;10:403–11.
[18] Vickers AJ, Elkin EB. Decision curve analysis: a novel method for evaluating pre- [40] Friberg J, Buch P, Scharling H, Gadsbphioll N, Jensen GB. Rising rates of hospital
diction models. Medical decision making: an international journal of the Society for admissions for atrial fibrillation. Epidemiology 2003;14:666–72.
Medical Decision Making 2006;26:565–74.

You might also like