American Journal of Emergency Medicine (2011) 29, 1158-1162

Brief Report

Utility of routine thyroid-stimulating hormone determination in new-onset atrial fibrillation in the ED^?

Francesco Buccelletti MD?, Annarita Carroccia MD, Davide Marsiliani MD, Emanuele Gilardi MD, Nicolo Gentiloni Silveri MD, Francesco Franceschi MD

Department of Emergency Medicine, Catholic University of the Sacred Heart, 00168 Rome, Italy

Received 18 May 2010; revised 7 June 2010; accepted 8 June 2010

Abstract

Introduction: Hyperthyroidism is a relative uncommon but important cause of atrial fibrillation. The aim of this study was to investigate the utility of routine thyroid-stimulating hormone (TSH) determination in the emergency department (ED) in patients presenting to the ED with stable, new-onset atrial fibrillation. We derive a set of clinical criteria in which TSH is likely to be normal and therefore thyroid function evaluation deferrable to a different time from ED visit.

Methods: Cross-sectional observational study in a university hospital. Thyroid-stimulating hormone was measured in all patients admitted to the ED observational unit for new-onset atrial fibrillation in a 30 consecutive months’ period. Patients’ clinical characteristics and treatment received in the ED were recorded. Recursive partitioning analysis technique was used to determine which predictors were associated with a TSH level less than 0.35 uIU/mL.

Results: Of 433 patients enrolled, 47 (10.8%) had a low TSH. Thyroid-stimulating hormone highly correlated with FT3 and FT4 levels (P b .001) confirming its good predictive value as screening tool. Recursive partitioning analysis showed that previous thyroid disease (P b .01), stroke/transient ischemic attack (P b .01), and hypertension (P = .10) were associated with low TSH. The final model had sensitivity of 93% and specificity of 31%, corresponding to a negative likelihood ratio of 0.02

(0.01-0.07).

Conclusion: Hyperthyroidism is present in nearly 10% of new-onset atrial fibrillation. Although thyroid function screening is recommended in all patients, a simple model that included previous thyroid disease, stroke, and hypertension might help to identify those patients at high risk (low TSH) in the ED.

(C) 2011

Introduction

Atrial fibrillation is the most common cardiac arrhythmia diagnosed and treated in an emergency depart-

^? Conflict of interest. Authors have no conflict of interest to declare.

* Corresponding author. Tel.: +39 06 30155394.

E-mail address: [email protected] (F. Buccelletti).

ment (ED). Atrial fibrillation occurs in 10% to 15% of patients with hyperthyroidism [1], and it is more common in men and with old age. Atrial fibrillation in thyrotoxicosis is associated with significant mortality and morbidity resulting from embolic events [2]. International guidelines [3] recommend that thyroid hormones (thyroid-stimulating hormone [TSH], free T3, and free T4) must be part of the evaluation in all new-onset AF without specifying whether this determination should be done in the ED on patient

0735-6757/$ - see front matter (C) 2011 doi:10.1016/j.ajem.2010.06.010

presentation. On the other hand, patients with AF and hyperthyroidism are known to poorly respond to cardiover- sion and a Rate control strategy would be the best choice.

Thyroid-stimulating hormone can be used as first line screening tool for thyroid function. When TSH is within normal limits, it virtually rules out thyroid hyperfunction [3]. There are few data on clinical utility of TSH determina- tion in the ED. Ideally, all patients with new-onset AF should be screened for secondary causes (including hyperthyroid- ism), but in reality, this screening, performed in emergency, increases costs, length of staying in the ED, and ultimately,

delays patient’s treatment.

The aim of this cross-sectional study was to develop a set of clinical criteria that could be used in the ED to identify AF patients who have a very low probability of abnormal (low) TSH and, therefore, a set of patients in which thyroid function evaluation can be deferred. A low level of TSH (b0.35 uIU/mL ) is considered a red flag for further evaluation for thyroid function and should call for free fraction hormones determination, early anticoagula- tion, and specific treatment along with arrhythmia control (ie, steroids) [4].

Methods

This was a cross-sectional observational study, performed in a tertiary care, university hospital, involving patients presenting to the ED and admitted to the ED observational unit, with new-onset AF between April 1, 2007, and December, 31, 2009. Exclusion criteria were hemodynamic unstable conditions requiring resuscitation at presentation and/or immediate electrical cardioversion, other supraven- tricular arrhythmia (ie, atrial flutter) on initial 12-lead electrocardiogram , age less than 18 years old, pregnancy, and home therapy with amiodarone. This study was not interventional; no new test was performed, and therefore was waived by ethical committee and approved by the department board of directors.

TSH, free T3, and free T4 levels were dosed by hospital central laboratory according to the hospital quality and standards. Blood was drawn upon patient admission to the ED observational unit as part of a standardized nursing protocol. Median time from ED triage to TSH determination was 3.5 hours (interquartile range, 2.1-5.7 hours). Normal ranges indicated by the laboratory were 0.35 to 2.80 uIU/mL for TSH, 2.3 to 4.2 pg/mL for free T3, and 8.05 to 15.05 pg/mL. Patient’s clinical characteristics were extracted from the electronic clinical chart (GIPSE(R), Agency for Public Health, Rome-Italy) that was used by ED physicians throughout the study duration, as per ED attending physician

notes (history and physical examination).

Initial 12-lead ECGs were reviewed by authors to confirm AF. Electrocardiograms were also read by a cardiologist as standard hospital policy.

Data analysis

Variables are summarized as absolute number and percentage when categorical and median and range when continuous. Univariate analysis between patients with TSH less than 0.35 uIU/mL and equal or above 0.35 uIU/mL was performed using ?2 test. P values less than .05 were considered to be significant. Likelihood ratio, odds ratio, and their relative 95% confidential intervals (95% CI) were calculated [5].

Recursive partitioning analysis was used to assess clinical variables identifying patients with TSH less than

0.35 uIU/mL. Starting with the initial set of data, we repeatedly removed the data on patients who had the clinical variable with the lowest P value (by ?2 analysis) for the comparison between patients with low TSH and those with normal TSH, until there was no patients with low TSH or other significant Clinical predictors. The likelihood ratio (LR) indicates the likelihood of a low TSH in patients with the clinical variable in question as compared with the likelihood in patients without it. This technique was previously used in emergency medicine literature, and it is particularly suitable when sensitivity needs to be maximized (rule-out models) [6].

Results

A total of 433 patients were eligible for this study. Thyroid- stimulating hormone was abnormally low (b0.35 uIU/mL) in 47 (10.8%) patients, with 27 (6.2%) patients having a TSH less than 0.1 uIU/mL. Low TSH correlated with high free T3 (n = 8, P b .001) and free T4 (n = 17, P b .001). Odds ratio of a high free T3 and free T4, having a low TSH, were, respectively, 173.71 (20.30-1486.52) and 11.12 (4.61- 26.83), demonstrating a good predictive value of the TSH as screening tool.

Table 1 shows demographics and clinical characteristics of patients with normal and low TSH (b0.35 uIU/mL). As shown, at univariate analysis, single predictors of having low TSH level were as follows: history of thyroid disease, which included previous thyroid surgery, previous known hypo- or hyperthyroidism and/or Long-term treatment with synthetic thyroid hormones at home (P b .001); history of previous Stroke or TIA (P b .001); and hypertension (P = .055).

Rhythm control strategy (pharmacologic and/or electric cardioversion) was used in 20 (42.5%) with low TSH. This rate was similar in the group with normal TSH.

Amiodarone, a drug known to alter thyroid function, was used in 10 (21.3%) cases in the low-TSH group and in 106 (27.3%) in the normal-TSH group (P = .366).

Fig. 1 shows results of recursive partitioning analysis. Considering history of thyroid disease, history of previous stroke/TIA, and hypertension, 3 of 121 patients were misdiagnosed, corresponding to a sensitivity of 93% (95%

TSH b0.35 (n = 47)			TSH N0.35 (n = 386)		LR+	P
Age a	71	(64-77)	72	(62-78)	7.43	b.720
Male	21	(44.6)	202	(52.3)	1.01	b.313
Previous episodes of AF	22	(46.8)	190	(49.2)	0.09	b.755
History of thyroid disease b	23	(48.9)	36	(9.3)	45.23	b.001
History of Stroke/TIA	9	(19.1)	17	(4.4)	11.68	b.001
History of hypertension	38	(80.8)	259	(67.0)	3.98	b.055
Valvular disease	1	(2.1)	27	(6.9)	2.12	b.200
Rhythm control c	20	(42.5)	194	(50.2)	0.99	b.318
History of HF	5	(11.1)	55	(14.2)	0.48	b.499
Signs of HF d	12	(25.5)	97	(25.1)	0.00	b.952
Diabetes	7	(14.8)	52	(13.4)	0.07	b.788
Coronary artery disease	7	(14.8)	60	(15.5)	0.01	b.907
Data are expressed as number (percentage) except for age, which is indicated as median (interquartile range). LR+ indicates positive likelihood ratio; HF, heart failure. a LR+ is calculated for intervals of 10 years. b Thyroid disease indicates previous thyroid surgery, previous known hypo- or hyperthyroidism, and/or long-term treatment with synthetic thyroid hormones at home. c Rhythm control indicates patients who received Antiarrhythmic drugs for pharmacologic cardioversion attempt. d At least 2 of the following symptoms/signs at presentation: dyspnea at rest, peripheral edema, jugular veins distension, signs of pulmonary edema on chest x-ray.

CI, 81%-98%), a specificity of 31% (95% CI, 26%-35%), and a negative likelihood ratio of 0.02 (95% CI, 0.01-0.07). Fig. 2 shows the ROC curve of the final model obtained.

Table 1 Clinical characteristics of the study population

The area under the curve was 0.75 (P b .001).

Discussion

Thyroid disease and AF

Thyroid hormones exert their Cardiovascular effects either directly through nuclear thyroid receptors or indirectly by

Fig. 1 Recursive partitioning analysis (TSH model).

influencing sympathoadrenergic system and altering periph- eral vascular resistance. These effects result in increased heart rate, systolic hypertension, increased ventricular contractility, and cardiac hypertrophy. In addition, subclin- ical hyperthyroidism is a risk factor that is associated with a 3-fold increase in risk of developing AF [1].

Fig. 2 Receiver operating characteristic curve of the recursive partitioning model. The area under the curve is 0.74 with 95% CI of 0.67-0.82 (P b .001).

Subclinical hyperthyroidism is defined as low serum thyrotropin concentration in asymptomatic patient with normal serum T3 and T4 concentration. It has a prevalence of 0.5% to 3.9% in adults [7] and 11.8% in the elderly [8]. Subclinical hyperthyroidism can occur as a result of thyroid pathology such as Graves’ disease, multinodular goiter, or autonomous toxic nodules or may be exogenous due to thyroxine therapy [8]. There is no significant difference in AF occurrence between overt and subclinical hyperthyroidism [1].

Electrocardiogram may be helpful in identifying hyper- thyroid subjects at risk for developing AF. Maximum P-wave duration and P-wave dispersion are higher in both subclinical and overt hyperthyroidism. P maximum and P wave dispersion are significant predictors of paroxysmal AF [9]. On the other hand, there is no ECG-specific feature in patients presenting for AF with underlying hyperthyroidism. The TSH-producing cells of the anterior pituitary gland are sensitive to minor changes in circulating thyroid hormones, and absent or subnormal TSH concentrations may be found in hyperthyroid patients in whom the T3 and T4 concentra- tions are higher than normal [10].

Mainstay of treatment in patients with AF and hyperthyroidism is restoration of euthyroid status. More than 50% of AF spontaneously reverts to sinus rhythm when the thyroid hormone levels start to decline. Second, it was previously noted that these patients are at higher risk of cardioembolic events compared to a similar population with AF but without hyperthyroidism. Guidelines recommend that these patients should be anticoagulated until euthyroid state is restored and [3].

ventricular rate control in patients with AF and hyperthyroidism should be achieved using preferably ?-blockers. Alternatively, when ?-blockers cannot be used, a calcium channel blocker, such altiazem, diltiazem or verapamil, might be chosen.

Utility of TSH in the ED

There are no previous studies investigating clinical utility of routine use of TSH in AF patients in the ED. Currently, because hyperthyroidism has low prevalence in AF, TSH is dosed when hyperthyroidism is clinically suspected. Even this practice is not evidence-based and missing hyperthy- roidism might lead to clinical strategy error in terms of rhythm vs rate control.

The goal of this study was to develop a simple set of clinical criteria identifying AF patients who have TSH level abnormally low in the ED. In this sample, TSH had a low but significant prevalence (10%), same as previously described in a larger cohort [7]. Patients with low TSH were initially treated with rhythm or rate control in similar proportion, indirectly indicating the low sensitivity of ED physicians to detect thyroid abnormality with clinical judgment. It is true that TSH does not imply thyrotoxicosis per se, but it is enough to call for further thyroid function evaluation and

maybe to opt initially for rate control strategy (which was not in this case).

Designing this study, recursive partitioning analysis was preferred to other statistical methods because of its ability to maximize sensitivity and its simplicity to interpret. Because thyroid disease has a low prevalence, whereas AF has a high prevalence in the ED, this model is designed to maximize resource utilization and costs in the ED. Recursive partitioning analysis indicated that in the absence of history of thyroid disease, previous stroke/TIA, and hypertension, a low TSH level is very unlikely. Starting with 433 TSH dosages, 121 (27.9%) of them were normal. Following the model, a consistent amount of TSH (almost 30%) might be saved in initial evaluation. Of 47 patients with low TSH, 3 (6.3%) were misdiagnosed and one of them had a TSH less than 0.1 uIU/mL and abnormal (high) free T4.

Thyroid, AF, and stroke

Guidelines recommend that patients with AF and hyperthyroidism should be anticoagulated until thyroid function is normalized and afterward according to the general AF population criteria. This recommendation is based on expert opinion [11]. The rationale is that patients presenting with AF in the setting of hyperthyroidism are more likely to have other episodes and therefore prone to have cardioembolic events.

In this study, one of the predictors of having an abnormal TSH, and actually the strongest after the obvious history of previous thyroid disease, was a (previous) stroke or TIA, providing an indirect evidence of the link between embolic event and thyroid disease in AF. It is not clear whether thyroid disease acts as embolic risk factor beyond the simple probability of AF recurrence.

Limitations

This study was conducted in a single hospital (in a single unit) and its conclusions might not be easily generalized to other clinical settings. The model presented cannot be applied to other contexts in which AF and thyroid disease prevalence are different, in particular out of the ED (eg, cardiology or ambulatory care), unless with further validation.

This study focused on TSH, and complete data on thyroid function (such thyroid ultrasound study) were not available. Thyroid-stimulating hormone was dosed upon patient admission to the observational unit. We relied on the assumption that a significant change in TSH level between

Triage time and TSH determination was very unlikely.

Conclusion

Hyperthyroidism has a relatively low, but significant, prevalence in patients presenting to the ED with AF.

Identifying these patients is particularly important in terms of AF management (anticoagulation and rate control until euthyroid state is restored). Thyroid-stimulating hormone is a powerful tool used to screen patients for hyperthyroidism. Using recursive partitioning analysis technique, in a relative large cohort in the ED observational unit, history of thyroid disease, previous stroke/TIA, and hypertension helped to identify patients at high risk for hyperthyroidism with a sensitivity of 93%.

This model might be used to better allocate resource in the ED, dosing TSH immediately in patients who have a higher pretest probability of thyroid disease and deferring the screening in the others.

References

1. Sawin CT, Geller A, Wolf P, et al. Low serum thyrotropin concentrations as a risk factor for atrial fibrillation inolder persons. N Engl J Med 1994;331:1249-52.
2. Staffurt JS, Gibberd JS, Tang FS. arterial embolism in thyrotoxicosis with atrial fibrillation. Br Med J 1977;2:688-90.
3. Fuster V, et al. ACC/AHA/ESC guidelines for the management of patients with atrial fibrillation. Circulation 2001;104:2118-50.
4. Goichot B, Sapin R, Schlienger JL. Subclinical hyperthyroidism: considerations in defining the lower limit of the thyrotropin Reference interval. Clin Chem 2009;55(3):420-4 [Epub 2009 Jan 15].
5. Weingart S, Wyer P. Emergency medicine decision making: critical issues in chaotic environments. (Columbus, OH 43272): The McGraw- Hill Company; 2006. p. 11-79.
6. Haydel MJ, Mills TJ, Luber S, et al. Indications for computed tomography in patients with Minor head injury. N Engl J Med 2000;343:100-5.
7. Bagchi N, Brown T, Parish R. Thyroid dysfunction in adults over age

55 years. A study in anurban US community. Arch Intern Med 1990;150:785-7.

1. Parle J, Franklyn J, Cross K, Jones C. Prevalence and follow-up of abnormal thyrotropin concentrations in the elderly in the United Kingdom. Clin Endocrinol 1991;34:77-83.
2. Aras D, Maden O, Ozdemir O, Aras S, et al. Simple electrocardio- graphic markers for the prediction of Paroxysmal atrial fibrillation in hyperthyroidism. Int J Cardiol 2005;99(1):59-64.
3. Forfar JC, Feek CM, Miller HC, et al. Atrial fibrillation and isolated suppression of the pituitary-thyroid axis: response to specific antithyroid therapy. Int J Cardiol 1981;1:43-8.
4. Andrew D, Krahn MD, et al. How useful is Thyroid function testing in patients with recent-onset atrial fibrillation? Arch Intern Med 1996;156(19):2221-4.