ABSTRACT
Context
Treating overt hyperthyroidism prevents atrial fibrillation (AF). Though subclinical hyperthyroidism (SH) has been associated with AF, it is unknown whether treating SH prevents AF.
Objective
We aimed to identify the association between treating SH and incident AF.
Design
In a pharmacoepidemiologic retrospective cohort study, patients diagnosed with SH between 2000 and 2021 were followed.
Patients
Outpatients ≥ 18 years with biochemical SH and without prior AF, hypothyroidism, thyroid cancer, pituitary disease, or pregnancy were included.
Main Outcomes
The primary outcome was incident AF. Secondary outcomes were ECG and echocardiographic features associated with AF.
Results
Of 2169 patients screened, 360 (131 treated and 229 untreated) were followed up for a mean of 4.27 years. In the treated and untreated groups, AF occurred in 4 (3.1%) and 15 (6.6%) patients (p = 0.15), and AF incidence was 0.8% and 1.4%/year (p = 0.31), respectively. The hazard ratio (HR) for treatment as a time‐dependent variable was 0.60 (95% CI 0.19–1.92; p = 0.39). As some cases of AF were documented nearly simultaneously with SH treatment, a sensitivity analysis was performed reassigning two patients diagnosed with AF < 30 days after starting SH treatment to the untreated group. Here, in the treated and untreated groups, AF occurred in 1.6% and 7.4% (p = 0.02), and AF incidence was 0.4% and 1.8%/year (p = 0.02), respectively. The HR was 0.25 (0.06–1.13; p = 0.07). There were no differences in ECG or echocardiographic features.
Conclusion
There was an overall trend towards lower incidence and prevalence of AF following treatment of SH, supporting the need for larger scale studies.
Keywords: atrial fibrillation, subclinical hyperthyroidism, treatment
1. Introduction
Atrial fibrillation (AF) is a common cardiac arrhythmia, with a worldwide prevalence of 59.7 million cases. The lifetime risk of developing AF in Europe and the United States is 25% for those over the age of 55 [1]. Overt hyperthyroidism is an established risk factor for AF [2, 3] with a relative risk of 6.0 [2, 3, 4]. Treatment of hyperthyroidism prevents the onset of AF (RRR = 31%) and is associated with the resolution of AF (RRR = 80%) [5, 6].
Subclinical hyperthyroidism (SH), defined as a low TSH with normal free T4 (FT4) and free T3 (FT3), has also been associated with AF in several studies [4, 7, 8, 9, 10]. The prevalence of SH ranges between 1% and 10% worldwide and is highest in iodine‐deficient countries [11]. The mean incidence of AF in those with SH is 1.87%/year [12]. SH is associated with increased P wave duration on electrocardiography (ECG), as well as increased ascending aorta and end‐diastolic ventricular diameters on echocardiogram [13, 14]. P‐wave duration and end‐diastolic left ventricular diameter are also associated with AF.
Very few studies have assessed the effect of treating SH on AF, with the available studies predominantly investigating the role of treating SH in AF resolution as well as in improving ECG and echocardiographic features associated with SH [4, 13, 15, 16]. The objective of our study was to identify whether an association exists between treatment of SH in outpatients and preventing incident AF.
2. Materials and Methods
2.1. Study Design
This was a retrospective cohort study. Patients with SH (defined as low TSH with normal free T4 [FT4] and free T3 [FT3]) between September 2000 and December 2021 were identified using the Hamilton Regional Laboratory Medicine Program database. A single value was used to identify those with SH in this epidemiologic analysis. Patients were screened for inclusion using the electronic medical record (EMR) systems at two hospital systems in Hamilton, Canada. Subsequently, data were collected using a standardized data collection form between 2022 and 2023. Ethics approval was obtained from the Hamilton Integrated Research Ethics Board (Project ID: 14973).
3. Study Population
Individuals aged 18 years and above entered the cohort at the time of first biochemical evidence of SH, measured in the outpatient setting. Individuals were excluded from analysis for the following reasons: (1) diagnosis of persistent, permanent, or paroxysmal AF before cohort entry; (2) history of hypothyroidism or thyroxine use; (3) pituitary disease; (4) treatment with antithyroid medications, radioactive iodine ablation, or thyroid surgery before cohort; (5) prior or current diagnosis of thyroid cancer; (6) SH detected during pregnancy; (7) use of non‐thyroid drugs which affect thyroid function, including clozapine, cladribine, dipyrone, carbamazepine, eliglustat, sodium iodide, I‐131, fexinidazole and macimorelin; (8) SH secondary to euthyroid sick syndrome (if 1—explicitly stated in clinical notes or 2—after excluding other causes of SH, the patient was described to have a type of illness commonly associated with sick euthyroid syndrome, such as infection, at the time of the thyroid function tests completion); and (9) insufficient information: these patients’ charts did not include sufficient information to allow for confirmation of inclusion criteria or draw a reasonable conclusion on whether treatment of SH was administered. Two independent members of the team assessed each participant's eligibility criteria and resolved disagreements through discussion.
3.1. Exposure, Control and Outcome
Exposure (treatment) was defined as the treatment of SH from the first documented date of antithyroid medications (methimazole or propylthiouracil), radioactive iodine ablation, or surgical thyroidectomy regardless of whether a biochemical euthyroid state was established. Controls were those who had no documentation of treatment for SH for the duration of follow‐up (untreated). Participants were categorized into treated and untreated groups based on whether they were ever treated for SH. In addition, follow‐up time was categorized as treated time or untreated time.
The primary outcome was incident AF. Secondary outcomes were (1) P wave duration on ECG and (2) end‐diastolic ventricular and ascending aorta sizes on echocardiogram.
3.2. Clinical Data
Baseline demographics included age and sex. We also collected data on conditions associated with AF and clinical factors which may impact the decisions related to treating SH. These included previous coronary artery disease, congestive heart failure, valvular heart disease (collectively referred to as cardiac disease), hypertension, diabetes, dyslipidemia, obstructive sleep apnoea, pulmonary embolism, chronic renal insufficiency, liver disease, smoking history, excess alcohol use (more than 14 drinks/week for men and 10 drinks/week for women), osteoporosis and post‐menopausal status [7, 17, 18, 19, 20, 21]. These data were collected using the available clinical notes. The causes of SH were recorded based on the clinical notes, positive TSH receptor antibody (TRAB) results (suggestive of Grave's disease), and findings of available radioactive iodine uptake (RAIU) and scan reports. Where a discrepancy existed, the clinical note was prioritized. The aetiology was labelled as ‘unknown’ if the clinical notes did not specify the cause, TRAB was negative or unavailable, and the RAIU and scan was unavailable or unrevealing. Finally, the SH treatment modalities were identified using the clinical notes.
3.3. Biochemical Data
Information regarding serum TSH, FT4 and FT3 at the first biochemical evidence of SH were collected. Where available, TRAB levels were recorded as either positive or negative. As laboratory instruments and reference intervals varied over time, SH diagnoses were based on reference intervals concurrent with the measurements.
3.4. Cardiac Testing Data
Diagnosis of AF was identified by ECG, echocardiography, ambulatory cardiac monitoring devices, or documentation of AF in the clinical notes. Where available, heart rate and P wave duration on all ECGs, as well as end‐diastolic ventricular and ascending aorta diameters on all echocardiograms were abstracted after SH diagnosis and after SH treatment start dates in the untreated and treated groups, respectively. P wave duration was measured manually using a magnifying glass and was defined as the distance between the initial deflection and the immediate next area where the isoelectric line is crossed again. Further, information regarding the ascending aorta and end‐diastolic ventricular diameters was extracted from the echocardiogram reports. The average of P wave duration, ascending aorta diameter and end‐diastolic ventricular diameter were calculated for each participant.
3.5. Statistical Analysis
Baseline characteristics and SH etiologies were compared between the treated and untreated groups, using Student's t‐test or chi‐square test of independence. Total number of incident cases of AF was compared in the treated and untreated groups by chi‐square test. Incidence rates for AF were calculated from the time of SH treatment in the treated group and from the time of SH diagnosis in the untreated group, and they were compared by the incidence rate ratio test. Pharmacoepidemiologic analysis was completed using Cox proportional hazards regression with treatment as a time‐dependent variable (counting pretreatment person‐time in the treated group as untreated), adjusted for age and hypertension. The probability of AF development was estimated using an extended Kaplan–Meier curve with treatment as a time‐dependent variable (i.e., an extended Kaplan–Meier curve does not represent fixed cohorts of patients but allows patients to contribute to different curves at different times during follow‐up depending on their exposure at the time (untreated or treated) [22]. The curves were compared using the log‐rank test. ECG and echocardiographic features were compared using Student's t‐test. Ninety‐five percent confidence intervals were calculated and a p value of less than 0.05 was considered significant. Data were analysed using SPSS version 29 and Microsoft Excel.
3.6. Sensitivity and Subgroup Analysis
To address the observation that some cases of AF were documented nearly simultaneously with treatment of SH, a post hoc sensitivity analysis was conducted, whereby patients diagnosed with AF less than 30 days after the initiation of treatment for SH were included in the untreated group. The rationale for this analysis is twofold: (1) to avoid a possible unfair evaluation of effect, considering that SH treatment may take several weeks to reduce the risk of AF and (2) to address possible misclassification of SH treatment or AF at a precise point in time, considering that medical records data may not precisely reflect timing of (a) treatment (which could be later than timing of documented prescription of treatment) or (b) AF onset (which likely precedes its documentation and could potentially influence the decision to treat SH).
As lower TSH, higher FT4 and higher FT3 have been associated with higher risk of AF [23], we conducted three subgroup analyses with participants separated into (1) TSH less than or equal to 0.1 mU/L and TSH greater than 0.1 mU/L, (2) FT4 greater than or equal to 15 pmol/L and FT4 less than 15 pmol/L and (3) FT3 greater than or equal to 5 pmol/L and FT3 less than 5 pmol/L. The TSH cutoff of 0.1 mU/L is chosen as the major guidelines recommend SH treatment if TSH < 0.1 mU/L [7, 20] The respective FT4 and FT3 cutoffs of 15 and 5 pmol/L are chosen as they represent the mid‐normal range in most assays. Linear regression was used to identify the presence of an interaction between treatment status and FT4 levels, where FT4 was treated as a continuous variable.
4. Results
4.1. Participant Characteristics
A total of 2169 individuals with biochemical SH were screened and 1809 were excluded, leaving 360 for analysis (Figure 1). There were 131 treated and 229 untreated participants. The baseline clinical and biochemical characteristics are shown in Table 1. There were no significant differences in age, sex, heart disease, hypertension and smoking between the treated and untreated groups. Osteoporosis was more frequent in the treated (23%) compared to the untreated group (15%) (p = 0.02). TSH was significantly lower in the treated group (0.1 ± 0.11 mU/L) versus the untreated group (0.15 ± 0.12 mU/L) (p < 0.01). Follow‐up time in the treated group included 185 person‐years of untreated time before treatment (mean 1.42 years) and 496 person‐years after treatment (mean 3.79 years). The untreated group had a total of 1039 person‐years of untreated follow‐up time (mean 4.54 years).
Figure 1.
Participant flow. Each patient was excluded for only one reason. The first exclusion criteria listed were applied.
Table 1.
Patient characteristics.
| Characteristic | Treated group (n = 131) | Untreated group (n = 229) | p‐value |
|---|---|---|---|
| Age (years) | 55.05 ± 17.88 | 55.27 ± 18.08 | 0.91 |
| Sex (male) | 25 (19%) | 54 (24%) | 0.321 |
| Diabetes | 26 (20%) | 39 (17%) | 0.52 |
| Dyslipidemia | 27 (21%) | 51 (22%) | 0.70 |
| Heart diseasea | 23 (18%) | 40 (18%) | 0.98 |
| OSA | 4 (3%) | 17 (7%) | 0.09 |
| PE | 3 (2%) | 7 (3%) | 0.67 |
| CKD | 7 (5%) | 14 (6%) | 0.76 |
| Family history of AF | 1 (1%) | 0 (0%) | 0.19 |
| Hypertension | 64 (49%) | 92 (41%) | 0.13 |
| Liver disease | 1 (1%) | 5 (2%) | 0.31 |
| Osteoporosis | 30 (23%) | 31 (14%) | 0.02 |
| Smoking | 38 (29%) | 59 (26%) | 0.52 |
| Excess alcohol | 3 (2%) | 7 (3%) | 0.68 |
| Post‐menopause | 50 (38%) | 78 (35%) | 0.55 |
| TSH (mU/L) | 0.10 ± 0.11 | 0.15 ± 0.12 | < 0.0001 |
| TSH ≤ 0.1 mU/L | 79 (60%) | 89 (39%) | 0.0001 |
| Free T4 (pmol/L) | 15.06 ± 3.28 | 14.46 ± 3.11 | 0.08 |
| Free T4 ≥ 15 pmol/L | 58 (44%) | 96 (42%) | 0.66 |
| Aetiologies of subclinical hyperthyroidism and treatment | |||||
|---|---|---|---|---|---|
| Treated group (n = 131) | Untreated group (n = 229) | p‐value | Treated with AF (n = 4) | Untreated with AF (n = 15) | |
| Aetiology | |||||
| Toxic noduleb | 79 (60%) | 75 (33%) | < 0.0001 | 4 (100%) | 7 (47%) |
| Graves’ disease | 36 (27%) | 16 (7%) | < 0.0001 | 0 | 0 |
| Immunotherapy | 2 (2%) | 23 (10%) | < 0.0001 | 0 | 1 (6%) |
| Thyroiditis | 1 (1%) | 15 (7%) | < 0.0001 | 0 | 0 |
| Other | 0 (0%) | 4 (2%) | < 0.0001 | 0 | 0 |
| Unknown | 13 (10%) | 96 (42%) | < 0.0001 | 0 | 7 (47%) |
| Treatment modalitiesc | |||||
| Antithyroid medications | 97 (74%) | 4 (100%) | |||
| Surgery | 42 (32%) | 0 | |||
| Radio‐Iodine ablation | 26 (20%) | 0 | |||
Note: Values are mean ± SD or n (%).
Abbreviations: AF, atrial fibrillation; CKD, chronic kidney disease; OSA, obstructive sleep apnoea; PE, pulmonary embolism.
CAD, heart failure, or valvular disease.
Combined toxic adenoma and multinodular goitre.
Participants may have had more than one treatment modality.
4.2. Aetiologies and Treatment Modalities
Toxic nodule (toxic adenoma and multinodular goitre combined) was the most commonly documented aetiology of SH and occurred in 60% and 33% of treated and untreated participants, respectively (p < 0.01) (Table 1). Grave's disease was documented in 27% of treated and 7% of untreated participants (p < 0.01). Amongst those who developed AF, 58% had toxic nodule(s) and no participant had Grave's disease. The treatment modalities used were antithyroid medications (74%), surgery (32%) and radioactive ablation (20%), with some participants receiving more than one treatment modality. All four treated patients who developed AF only received antithyroid medications. The aetiology of SH was unknown in 13 (10%) and 96 (42%) of patients in the treated and untreated groups, respectively (Table 1). Out of these patients, 2 and 29 patients had negative TRAB in the treated and untreated groups, respectively.
4.3. Primary Outcome
AF occurred in 4 (3.1%) and 15 (6.6%) of treated and untreated participants, respectively (p = 0.15) (Table 2). The AF incidence rates in the treated and untreated groups were 0.8% and 1.4%/year, respectively (IRR 0.56, 95% CI 0.14–1.76; p = 0.31). In a time‐dependent Cox model, the adjusted hazard ratio (HR) for treatment as a time‐dependent variable was 0.60 (95% CI 0.19–1.92; p = 0.39) (Table 2). Extended Kaplan–Meier curve demonstrated a lower probability of AF development with treatment, most pronounced between 8 and 12 years after diagnosis of SH (Figure 2).
Table 2.
Association between treatment of subclinical hyperthyroidism and incident atrial fibrillation.
| Atrial fibrillation | Treated SH | Untreated SH | Effect size (95% CI) | p‐value |
|---|---|---|---|---|
| Probability | n (%) | n (%) | OR | |
| Overall | 4 (3%) | 15 (7%) | 0.45 (0.15–1.38) | 0.15 |
| Subgroups | ||||
| TSH ≤ 0.1 mU/L | 2 (3%) | 5 (6%) | 0.44 (0.08–2.32) | 0.32 |
| TSH > 0.1 mU/L | 2 (4%) | 10 (7%) | 0.55 (0.12–2.61) | 0.40 |
| FT4 ≥ 15.0 pmol/L | 2 (3%) | 8 (8%) | 0.39 (0.08–1.92) | 0.20 |
| FT4 < 15.0 pmol/L | 2 (3%) | 7 (5%) | 0.51 (0.10–2.51) | 0.40 |
| FT3 ≥ 5.0 pmol/L | 4 (4%) | 6 (5%) | 0.91 | 0.89 |
| FT3 < 5.0 pmol/L | 0 (0%) | 9 (9%) | N/A | 0.06 |
| Sensitivity (treated ≥ 30 days)a | 2 (2%) | 17 (7%) | 0.19 (0.04–0.85) | 0.01 |
| Incidence rate b | (%/year) | (%/year) | IRR | |
|---|---|---|---|---|
| Overall | 0.8 | 1.4 | 0.56 (0.14–1.76) | 0.31 |
| Subgroups | ||||
| TSH ≤ 0.1 mU/L | 0.6 | 1.6 | 0.38 (0.04–2.38) | 0.27 |
| TSH > 0.1 mU/L | 1.3 | 1.4 | 0.93 (0.10–4.39) | 0.99 |
| FT4 ≥ 15.0 pmol/L | 0.7 | 2.0 | 0.40 (0.04–1.98) | 0.24 |
| FT4 < 15.0 pmol/L | 0.9 | 1.4 | 0.64 (0.06–3.52) | 0.67 |
| FT3 ≥ 5.0 pmol/L | 1.0 | 1.2 | 0.84 (0.17–3.51) | 0.80 |
| FT3 < 5.0 pmol/L | 0.0 | 2.0 | N/A | N/A |
| Sensitivity (Treated ≥ 30 days)a | 0.4 | 1.8 | 0.22 (0.03–0.93) | 0.02 |
| Cox Hazardc | HR | |
|---|---|---|
| Overall | 0.60 (0.19–1.92) | 0.39 |
| Subgroups | ||
| TSH ≤ 0.1 mU/L | 0.35 (0.05–2.39) | 0.28 |
| TSH > 0.1 mU/L | 0.72 (0.14–3.64) | 0.70 |
| FT4 ≥ 15.0 pmol/L | 0.19 (0.02–1.59) | 0.12 |
| FT4 < 15.0 pmol/L | 0.29 (0.34–2.61) | 0.27 |
| FT3 ≥ 5.0 pmol/L | 1.26 (0.30–5.31) | 0.75 |
| FT3 < 5.0 pmol/L | N/A | N/A |
| Sensitivity (treated ≥ 30 days)a | 0.25 (0.06–1.13) | 0.07 |
In sensitivity analysis, patients diagnosed with AF < 30 days after starting treatment for SH were included in the untreated group.
For the treated group, follow‐up is from the start of SH treatment to either AF or the last follow‐up. For the untreated group, follow‐up time is from SH diagnosis to either AF or last follow‐up.
Treatment as time‐dependent variable and adjusted for age and hypertension (95% CI).
Figure 2.
Extended Kaplan–Meier curve of the probability of atrial fibrillation (AF) development with treatment as a time‐dependent variable.
In a sensitivity analysis, whereby two participants diagnosed with AF < 30 days after starting SH treatments were reassigned to the untreated group, AF occurred in 2 (1.6%) treated and 17 (7.4%) untreated participants, (p = 0.02) (Table 3), and the AF incidence rates in the treated and untreated groups were 0.4% and 1.8%/year, respectively (IRR 0.22, 95% CI 0.03–0.93; p = 0.02). The adjusted HR was 0.25 (0.06–1.13; p = 0.07) (Table 2). Of note, these two participants were treated with antithyroid medications, specifically methimazole. They were post‐menopausal women who also had osteoporosis, coronary artery disease and congestive heart failure.
Table 3.
ECG and echocardiogram characteristics.
| ECG/echocardiogram characteristic | Untreateda | Treatedb | p‐value |
|---|---|---|---|
| P‐wave duration (ms) | 73.42 ± 12.54 | 73.47 ± 11.61 | 0.99 |
| Left end‐diastolic ventricular diameter (cm) | 4.34 ± 0.52 | 4.29 ± 0.70 | 0.86 |
| Aorta diameter (cm) | 3.23 ± 0.39 | 3.33 ± 0.34 | 0.58 |
| Heart rate | 78.59 ± 16.15 | 77.43 ± 16.13 | 0.76 |
Note: Values are mean ± SD.
Mean value of all ECGs and echocardiograms after the date of SH diagnosis.
For treated group, the mean values of all ECGs and echocardiograms included were after the initiation of treatment (medications, radio ablation, thyroidectomy).
4.4. Subgroup Analysis
Amongst those with TSH ≤ 0.1 mU/L, AF incidence rates in the treated and untreated groups were 0.6% and 1.6%/year, respectively (IRR 0.38, p = 0.27), whereas AF incidence rates amongst the participants with TSH > 0.1 mU/L in the treated and untreated groups were 1.3% and 1.4%/year, respectively (IRR 0.93, p = 0.99). The adjusted HR for treatment in those with TSH ≤ 0.1 mU/L and TSH > 0.1 mU/L were 0.35 (0.05–2.39; p = 0.28) and 0.72 (0.14–3.64; p = 0.70), respectively (Table 2).
Amongst those with FT4 ≥ 15 pmol/L, AF incidence rates in the treated and untreated groups were 0.7% and 2.0%/year, respectively (IRR 0.40, p = 0.24), while AF incidence rates amongst the treated and untreated groups with FT4 < 15 pmol/L were 0.9% and 1.4%/year, respectively (IRR 0.64, p = 0.67). The adjusted HR in those with FT4 ≥ 15 pmol/L and FT4 < 15 pmol/L were 0.19 (0.02–1.59; p = .12) and 0.29 (0.34–2.61; p = 0.27), respectively (Table 2). An interaction term between treatment and FT4, where FT4 was treated as a continuous variable, was not statistically significant. Finally, amongst those with FT3 ≥ 5 pmol/L, AF incidence rates in the treated and untreated groups were 1.0% and 1.2%/year, respectively, with an adjusted HR of 1.26 (0.30–5.31; p = 0.75). No treated patient with FT3 < 5 pmol/L developed AF and, therefore, the incidence and HR in the untreated groups were not calculated.
4.5. Secondary Outcomes
The mean P‐wave duration, end‐diastolic ventricular diameter and ascending aorta diameter were 73.47 versus 73.42 ms (p = 0.99), 4.29 versus 4.34 cm (p = 0.86), and 3.33 versus 3.23 cm (p = 0.58) in the treated and untreated groups, respectively (Table 3).
5. Discussion
We report the first study examining the effect of SH in preventing incident AF. Our findings show an overall trend towards a lower incidence of AF following SH treatment. Despite this trend, no statistical significance was observed. In a subsequent sensitivity analysis, where those diagnosed with AF < 30 days after starting treatment for SH were reassigned to the untreated group, the AF incidence rate was statistically significantly lower in the treated group. The lack of significant difference in the main group was likely related to the small sample size and the influence of new AF on treatment decisions. The respective decrease in AF frequency and incidence from 6.6% to 3.1% and 1.4% to 0.8%/year, following treatment in the main group would be considered clinically important if corroborated by larger studies.
Our results support the previous small studies demonstrating the resolution of AF following the treatment of SH amongst those with pre‐existing AF. Forfar et al. examined four patients with AF and SH, where initial cardioversion did not restore sinus rhythm in three of the patients. However, all four patients converted to and remained in sinus rhythm at a 2‐year follow‐up after treatment of SH and restoration of the euthyroid status [15]. In a cohort study, amongst 78 patients with SH and AF, 15 (19%) of those who achieved biochemical euthyroidism following treatment of SH converted to sinus rhythm. However, this study did not report how many, if any, untreated patients converted to sinus rhythm during the study period [4]. A randomized controlled trial in France aiming to assess the role of treating SH in preventing the rate of AF was terminated due to recruitment difficulties (clinicaltrials.gov identifier NCT00213720).
Previous studies demonstrated that restoration of euthyroid status following treatment of SH led to a decrease in ascending aorta diameter, end‐diastolic ventricular diameter and heart rate [13, 16]. Amongst patients with overt hyperthyroidism and AF, achieving euthyroid status following treatment of hyperthyroidism was associated with improvement in P wave duration [24]. Our study, however, did not demonstrate any changes in P wave duration, heart rate, ascending aorta diameter and end‐diastolic ventricular diameter following treatment of SH. Several methodological differences exist between these and our studies, which could explain these differences: (1) whereas the other studies assessed the ECG and echocardiographic features at a certain time point following restoring euthyroidism, achieving euthyroidism was not a part of the definition of treatment of SH in our study. (2) The three other studies followed one patient over time to compare the ECG and echocardiographic features before and after the treatment of SH, while we compared two groups of participants with SH with the intervention group receiving treatment for SH and the control group not receiving it; and (3) all of the other studies were prospective, allowing investigators to identify a consistent time to perform ECG and echocardiograms, however, we retrospectively analysed and averaged all ECGs and echocardiograms after SH treatment and diagnosis in the treated and untreated groups, respectively.
Most (58%) patients with AF and SH, and all treated patients with SH and AF had thyroid nodular disease. Our results support the findings by Turan et al., showing amongst a group of 36 patients with hyperthyroidism, AF rates were higher in those with toxic nodular goitre (TNG) (p = 0.02) [24]. The average age was higher in those with TNG. Given age is a risk factor for AF, the association between thyroid nodular disease and AF is largely related to age [25]. These findings suggest that the aetiology of SH should be considered when deciding on treatment.
All four treated patients who developed AF only received antithyroid medications, and all four had toxic nodules. This finding supports the recommendations by the current guidelines that surgical over medical treatment should be prioritized for individuals with toxic nodules [7, 20].
A larger proportion of patients were recorded as having an ‘unknown aetiology’ in the treated (42%) versus the untreated (10%) groups, respectively. This is an inherent limitation of retrospective chart reviews. A reason for the discrepancy in the proportion of unknown etiologies between the two groups may be that, to some extent, the aetiology of SH dictates the treatment modality.
In a subgroup analysis based on TSH and FT4, a stronger though nonsignificant trend towards lower incidence of AF was observed amongst participants with TSH ≤ 0.1 mU/L compared to those with TSH > 0.1 mU/L, and amongst participants with FT4 ≥ 15 pmol/L compared to those with FT4 < 15 pmol/L. The American Thyroid Association (ATA) and the European Thyroid Association (ETA) guidelines use a TSH < 0.1 as an indication for the treatment of SH [7, 20]. A systematic review and meta‐analysis showed that although TSH, FT4 and FT3 were associated with AF, the association with FT4 was stronger compared with TSH [23]. Our findings support the existing literature that both FT4 and TSH levels should be factored into treatment decisions. A similar comparison was not possible in the subgroup analysis involving FT3, as no treated patients with FT3 < 5 developed AF.
There are several strengths in our study. The use of Cox regression analysis with treatment as a time‐dependent variable avoided immortal time bias [26]. In addition, identifying patients from an outpatient biochemistry database avoided the inclusion of patients with sick euthyroid syndrome and minimized the possible selection bias inherent in a cardiology or endocrinology database. As a result, the rate of AF in our patients matched that of the Canadian population, and there were similar clinical baseline characteristics between the two groups [25]. Other than osteoporosis and TSH levels, no significant differences were present between the two groups. These differences can be explained, as ATA and ETA recommend considering both osteoporosis and TSH levels when deciding on the treatment of SH [7, 20]. The presence of osteoporosis is not expected to affect the primary or secondary outcomes of the study. The TSH levels, however, are associated with the study outcome. The differences in TSH between the two groups may pose a confounding bias, leading to a larger number of patients with AF in the treated group. Therefore, the TSH difference between the groups is considered a study limitation.
Additional limitations include the small sample size and small number of events despite screening over 2000 patients, the single‐centre and retrospective design. Reference ranges for thyroid function tests changed over time, creating heterogeneity, particularly posing challenges with our subgroup analysis based on thyroid function tests. Moreover, given our methodology followed an intention‐to‐treat approach, we did not examine whether and how long after the start of treatment euthyroidism was established in each participant. In clinical practice, those treated for hyperthyroidism or SH may not consistently develop and maintain euthyroidism. Avoiding the exclusion of those with abnormal thyroid function tests mimics the real clinical scenario more closely. Finally, insufficient clinical information was available for 420 of the patients with SH within the hospital EMRs. This may represent a selection bias as patients whose SH might have been addressed outside of the main academic hospitals were excluded.
Our findings though not conclusive, along with similar recently published literature, suggest that treatment of SH to prevent AF should be considered, particularly in those with AF risk factors. We also showed that patients with SH secondary to toxic nodules are more likely to develop AF and, therefore, the aetiology of SH should be considered when deciding on treatment. Finally, TSH and FT4 levels may be factored into decisions regarding SH treatment. Our results support the need for a larger prospective study, ideally a randomized trial, assessing the effect of SH treatment in preventing AF. Additional questions raised by this research include whether early versus delayed approach to treatment of SH would make a difference in the incidence of AF and whether unique patient and disease characteristics would allow for active surveillance of patients with SH with minimal risk of developing AF.
Conflicts of Interest
Peter Kavsak: Outside of this work, Dr. Kavsak has received grants/reagents/consultant/advisor/honoraria from Abbott Laboratories, Abbott Point of Care, Beckman Coulter, Ortho Clinical Diagnostics, Quidel, Randox Laboratories, Roche Diagnostics, Siemens Healthcare Diagnostics and Thermo Fisher Scientific. McMaster University has the following patent with Dr. Kavsak listed as an inventor: ‘Method of Determining Risk of an Adverse Cardiac Event’. The other authors declare no conflicts of interest.
References
- 1. Elliott A. D., Middeldorp M. E., Van Gelder I. C., Albert C. M., and Sanders P., “Epidemiology and Modifiable Risk Factors for Atrial Fibrillation,” Nature Reviews Cardiology 20, no. 6 (2023): 404–417. [DOI] [PubMed] [Google Scholar]
- 2. Frost L., Vestergaard P., and Mosekilde L., “Hyperthyroidism and Risk of Atrial Fibrillation or Flutter: A Population‐Based Study,” Archives of Internal Medicine 164, no. 15 (2004): 1675–1678. [DOI] [PubMed] [Google Scholar]
- 3. Rahman F., Kwan G. F., and Benjamin E. J., “Global Epidemiology of Atrial Fibrillation,” Nature Reviews Cardiology 11, no. 11 (2014): 639–654. [DOI] [PubMed] [Google Scholar]
- 4. Auer J., Scheibner P., Mische T., Langsteger W., Eber O., and Eber B., “Subclinical Hyperthyroidism as a Risk Factor for Atrial Fibrillation,” American Heart Journal 142, no. 5 (2001): 838–842. [DOI] [PubMed] [Google Scholar]
- 5. Lillevang‐Johansen M., Abrahamsen B., Jørgensen H. L., Brix T. H., and Hegedüs L., “Duration of Hyperthyroidism and Lack of Sufficient Treatment Are Associated With Increased Cardiovascular Risk,” Thyroid 29, no. 3 (2019): 332–340. [DOI] [PubMed] [Google Scholar]
- 6. Ryödi E., Metso S., Huhtala H., Välimäki M., Auvinen A., and Jaatinen P., “Cardiovascular Morbidity and Mortality After Treatment of Hyperthyroidism with Either Radioactive Iodine or Thyroidectomy,” Thyroid 28, no. 9 (2018): 1111–1120. [DOI] [PubMed] [Google Scholar]
- 7. Ross Douglas S., Burch Henry B., Cooper David S., et al., “American Thyroid Association Guidelines for Diagnosis and Management of Hyperthyroidism and Other Causes of Thyrotoxicosis,” Thyroid 26, no. 10 (2016): 1343–1421. [DOI] [PubMed] [Google Scholar]
- 8. Cappola A. R., Fried L. P., Arnold A. M., et al., “Thyroid Status, Cardiovascular Risk, and Mortality in Older Adults,” JAMA 295, no. 9 (2006): 1033–1041. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9. Collet T.‐H., Gussekloo J., Bauer D. C., et al., “Subclinical Hyperthyroidism and the Risk of Coronary Heart Disease and Mortality,” Archives of Internal Medicine 172, no. 10 (2012): 799–809. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Sawin C. T., Geller A., Wolf P. A., et al., “Low Serum Thyrotropin Concentrations as a Risk Factor for Atrial Fibrillation in Older Persons,” New England Journal of Medicine 331, no. 19 (1994): 1249–1252. [DOI] [PubMed] [Google Scholar]
- 11. Carlé A., Andersen S. L., Boelaert K., and Laurberg P., “Management of Endocrine Disease: Subclinical Thyrotoxicosis: Prevalence, Causes and Choice of Therapy,” European Journal of Endocrinology 176, no. 6 (2017): R325–R337. [DOI] [PubMed] [Google Scholar]
- 12. Delitala A., “Subclinical Hyperthyroidism and the Cardiovascular Disease,” Hormone and Metabolic Research 49, no. 10 (2017): 723–731. [DOI] [PubMed] [Google Scholar]
- 13. Kaminski G., Michalkiewicz D., Makowski K., et al., “Prospective Echocardiographic Evaluation of Patients with Endogenous Subclinical Hyperthyroidism and After Restoring Euthyroidism,” Clinical Endocrinology 74, no. 4 (2011): 501–507. [DOI] [PubMed] [Google Scholar]
- 14. Çetinarslan B., Akkoyun M., Cantürk Z., Tarkun I., Kahranman G., and Komsuoglu B., “Duration of the P Wave and P Wave Dispersion in Subclinical Hyperthyroidism,” Endocrine Practice 9, no. 3 (2003): 200–203. [DOI] [PubMed] [Google Scholar]
- 15. Forfar J. C., Feek C. M., Miller H. C., and Toft A. D., “Atrial Fibrillation and Isolated Suppression of the Pituitary‐Thyroid Axis: Response to Specific Antithyroid Therapy,” International Journal of Cardiology 1, no. 1 (1981): 43–48. [DOI] [PubMed] [Google Scholar]
- 16. Sgarbi J. A., Villaça F. G., Garbeline B., Villar H. E., and Romaldini J. H., “The Effects of Early Antithyroid Therapy for Endogenous Subclinical Hyperthyroidism in Clinical and Heart Abnormalities,” Journal of Clinical Endocrinology and Metabolism 88, no. 4 (2003): 1672–1677. [DOI] [PubMed] [Google Scholar]
- 17. Dai H., Zhang Q., Much A. A., et al., “Global, Regional, and National Prevalence, Incidence, Mortality, and Risk Factors for Atrial Fibrillation, 1990–2017: Results from the Global Burden of Disease Study 2017,” European Heart Journal—Quality of Care and Clinical Outcomes 7, no. 6 (2020): 574–582. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18. Lau D. H., Nattel S., Kalman J. M., and Sanders P., “Modifiable Risk Factors and Atrial Fibrillation,” Circulation 136, no. 6 (2017): 583–596. [DOI] [PubMed] [Google Scholar]
- 19. Benjamin E. J., “Independent Risk Factors for Atrial Fibrillation in a Population‐Based Cohort: The Framingham Heart Study,” JAMA 271, no. 11 (1994): 840–844. [PubMed] [Google Scholar]
- 20. Biondi B., Bartalena L., Cooper D. S., Hegedüs L., Laurberg P., and Kahaly G. J., “The 2015 European Thyroid Association Guidelines on Diagnosis and Treatment of Endogenous Subclinical Hyperthyroidism,” European Thyroid Journal 4, no. 3 (2015): 149–163. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 21. Brandes A., Smit M. D., Nguyen B. O., Rienstra M., and Van Gelder I. C., “Risk Factor Management in Atrial Fibrillation,” Arrhythmia and Electrophysiology Review 7, no. 2 (2018): 118–127. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22. Snapinn S. M., Jiang Q., and Iglewicz B., “Illustrating the Impact of a Time‐Varying Covariate with an Extended Kaplan–Meier Estimator,” American Statistician 59, no. 4 (2005): 301–307. [Google Scholar]
- 23. Fitzgerald S. P., Bean N. G., Falhammar H., and Tuke J., “Clinical Parameters Are More Likely to Be Associated With Thyroid Hormone Levels Than With Thyrotropin Levels: A Systematic Review and Meta‐Analysis,” Thyroid 30, no. 12 (2020): 1695–1709. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24. Turan E., “Comparison of Cardiac Arrhythmia Types Between Hyperthyroid Patients With Graves' Disease and Toxic Nodular Goiter,” Acta Endocrinologica (Bucharest) 14, no. 3 (2018): 324–329. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Allan G. M., Morros M. P., and Young J., “Subclinical Hypothyroidism and TSH Screening,” Canadian Family Physician [Medecin de famille canadien] 66, no. 3 (2020): 188. [PMC free article] [PubMed] [Google Scholar]
- 26. Jones M. and Fowler R., “Immortal Time Bias in Observational Studies of Time‐to‐Event Outcomes,” Journal of Critical Care 36 (2016): 195–199. [DOI] [PubMed] [Google Scholar]