This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR mhealth and uhealth, is properly cited. The complete bibliographic information, a link to the original publication on http://mhealth.jmir.org/, as well as this copyright and license information must be included.
Symptoms and signs of thyrotoxicosis are nonspecific and assessing its clinical status is difficult with conventional physical examinations and history taking. Increased heart rate (HR) is one of the easiest signs to quantify this, and current wearable devices can monitor HR.
We assessed the association between thyroid function and resting HR measured by a wearable activity tracker (WD-rHR) and evaluated the clinical feasibility of using this method in patients with thyrotoxicosis.
Thirty patients with thyrotoxicosis and 10 controls were included in the study. Participants were instructed to use the wearable activity tracker during the study period so that activity and HR data could be collected. The primary study outcomes were verification of changes in WD-rHR during thyrotoxicosis treatment and associations between WD-rHR and thyroid function. Linear and logistic model generalized estimating equation analyses were performed and the results were compared to conventionally obtained resting HR during clinic visits (on-site resting HR) and the Hyperthyroidism Symptom Scale.
WD-rHR was higher in thyrotoxic patients than in the control groups and decreased in association with improvement of thyrotoxicosis. A one standard deviation–increase of WD-rHR of about 11 beats per minute (bpm) was associated with the increase of serum free T4 levels (beta=.492, 95% CI 0.367-0.616,
Heart rate data measured by a wearable device showed reasonable predictability of thyroid function. This simple, easy-to-measure parameter is clinically feasible and has the potential to manage thyroid dysfunction.
ClinicalTrials.gov NCT03009357; https://clinicaltrials.gov/ct2/show/NCT03009357 (Archived by WebCite at http://www.webcitation.org/70h55Llyg)
Thyrotoxicosis is a clinical syndrome resulting from the high concentration of free thyroxine (T4) and free triiodothyronine (T3). The prevalence of thyrotoxicosis is approximately 2%, and its most common cause (ie, 60%-90%) is Graves’ disease (GD) which is an autoimmune disease that stimulates the thyroid gland to produce and release thyroid hormone [
The popularity of wearable activity trackers has grown considerably in recent years. The American College of Sports Medicine survey of fitness trends reported that wearable technology was the top-rated trend in 2016 [
Therefore, the objectives of the present study were to investigate whether and how well HR data collected by commercially available activity trackers reflect thyroid function across the clinical course of thyrotoxicosis. Using this approach, the clinical feasibility of wearable devices for the management of thyroid dysfunction was evaluated.
This was a single-center prospective observational study. Subjects were recruited from the outpatient clinic of the endocrinology department at Seoul National University Bundang Hospital (SNUBH).
For the thyrotoxicosis group, patients 15-60 years of age who had been diagnosed with newly developed or recurrent thyrotoxicosis were eligible to participate. Participants needed to own a mobile phone and to be able to use a wearable device and its mobile application. Among those for whom the etiology of thyrotoxicosis was GD, only patients who had planned treatment with antithyroid drugs (ATD) were included so that their clinical course could be followed during medical treatment. We prescribed methimazole as the first choice ATD unless the subjects were contraindicated. No subject had a contraindication for methimazole or adverse events during the administration of methimazole. Inclusion and exclusion criteria are listed in
Healthy adults without a history of thyroid disease were included in the control group. Consistent with the thyrotoxicosis group, participants needed to be able to use a wearable device and the mobile application. These participants were screened to ensure they were not taking medications affecting HR, including beta blockers. There were 13 potential participants screened, and 10 were enrolled in the control group. All participants were informed about the study and provided written informed consent. This study was approved by the SNUBH Institutional Review Board (IRB #B-1609-363-004) and registered on ClinicalTrials.gov (trial registration #NCT03009357).
The study design is shown in
Study design and flow. Blood tests included a thyroid function test, serum levels of antithyroid-stimulating hormone receptor antibody, and other biochemical tests. Tc-99m is used in the thyroid scan. ATD: antithyroid drug, HSS: Hyperthyroidism Symptom Scale, P/Ex: physical examination.
After a one- to two-week screening period, patients visited the clinic to confirm the results of the TFT and other tests to determine the etiology (eg, autoantibodies, thyroid scan) and to start appropriate treatment. At this visit, those who met the inclusion criteria were enrolled. Patients with GD were prescribed a specific ATD dose, as determined by the endocrinologist. Patients with thyrotoxicosis caused by thyroiditis were reassured that their symptoms and signs were benign and self-limited. Patients taking propranolol were instructed to take the medication when their symptoms were severe and to inform the investigator of their dosing times, which were recorded in the case report form (CRF). Regardless of the etiology of thyrotoxicosis, all patients had monthly follow-ups, during which time they underwent blood tests, including TFT. Their ATD dose was adjusted as necessary. The study ended after each patient’s third visit; however, the study duration could be extended at the discretion of the investigator with the patient’s consent if their TFT was not fully restored. At each visit, anthropometric data were collected, and vital signs were measured.
Healthy adults were recruited into the control group through an official SNUBH announcement. Control participants visited the hospital on the same schedule as the thyrotoxic patients. They were given the same instruction about using the device and mobile application, and they were also instructed to wear the device all day and to inform to the investigator of any medication changes, which were recorded in the CRF. Control participants’ visit schedules and blood tests were consistent with those of thyrotoxic patients, but the duration of their study participation was not extended.
We used the Fitbit Charge HR or Fitbit Charge 2 (Fitbit, San Francisco, CA) and the Fitbit application for iOS (Apple, Cupertino, CA) or Android (Google, Mountain View, CA). The firmware versions of these devices were 18.128 for Fitbit Charge HR and 22.53.4 for Fitbit Charge 2 at the end of the study, and the latest version was maintained continuously over the study period. Although we started the study with Fitbit Charge HR, when Fitbit released Fitbit Charge 2, they discontinued production of the former device. Therefore, we used Fitbit Charge 2 with study participants enrolled after March 2017 (n=7 in the control group). However, these 2 models share a common sensor and data processing algorithm for both activity tracking and HR measurement. Activity and HR data are collected by the 3-axis accelerometer and plethysmography sensor, respectively. These sensors are equipped with the device. During the study period, each participant’s Fitbit account information, including identification and password, were shared with researchers, allowing us to access their online Fitbit account [
To assess their hyperthyroidism clinical status at each clinic visit, the endocrinologist overseeing this study evaluated the patients using the Hyperthyroidism Symptom Scale (HSS) [
The American Heart Association defines the resting heart rate (rHR) as the heart beats per minute (bpm) pumping the lowest amount of blood someone needs when they are in a resting position. The rHR can be changed by emotional state, medication, or current disease. In this study, we focused on rHR because we expected thyrotoxicosis to affect rHR significantly. We used 3 different rHR parameters according to the measurement method and calculating algorithm as follows. On-site rHR is the heart rate measured manually on the right wrist (ie, radial pulse) in a seated position after at least 10 minutes of resting. To calculate rHR from the HR log generated by the wearable device, we downloaded daily summary and detailed HR and activity data from the Fitbit database in JavaScript object notation (JSON) format using the application programming interface provided by Fitbit [
We measured the subjects’ height and weight while wearing light clothing and without shoes to the nearest 0.1 cm and 0.1 kg, respectively. Body mass index was calculated by determining the ratio between weight and the square of the height and expressed in kilograms per square meter. Right arm blood pressure was measured with the subject in a seated position after at least 10 minutes of resting. Biochemical measurements including TFT are listed in the
Data were expressed as the mean (SD) or median (interquartile range). To compare variables between the patient and control groups, we used Student
Participant baseline characteristics are summarized in
In thyrotoxic patients, thyrotoxicosis improved with treatment, as shown by decreased serum free T4 levels through the third visit. Serum TSH levels remained unchanged during the observation period (
Baseline characteristics.
Characteristics | Thyrotoxicosis (n=28) | Control (n=10) | ||
Age (years), mean (SD) | 34.9 (10.9) | 34.1 (5.9) | .78a | |
Male | 10 (36) | 3 (30) | .53b | |
Female | 18 (64) | 7 (70) | ||
Body mass index (kg/m2), mean (SD) | 20.6 (4.7) | 20.7 (1.6) | .93a | |
Systolic blood pressure | 130.2 (14.3) | 124.6 (11.4) | .27a | |
Diastolic blood pressure | 78.1 (10.6) | 70.6 (9.6) | .06a | |
On-site resting heart rate (bpmc), mean (SD) | 101.6 (14.5) | 81.9 (14.8) | .001a | |
Hyperthyroid symptom scale, mean (SD) | 12.5 (10.0) | 0.5 (2.8) | <.001d | |
Free thyroxine (ng/dL), mean (SD) | 3.08 (1.09) | 1.36 (0.12) | <.001a | |
Thyroid stimulating hormone (mIU/L), median (IQRe) | 0.01 (0.00) | 1.33 (1.19) | <.001d | |
Thyrotropin-binding inhibitory immunoglobulin (IU/L), median (IQR) | 4.2 (10.3) | — | — | |
Glucose (mg/dL), mean (SD) | 105.8 (18.6) | 96.6 (23.1) | .24a | |
Blood urea nitrogen (mg/dL), mean (SD) | 13.0 (2.9) | 11.7 (2.4) | .20a | |
Creatinine (mg/dL), mean (SD) | 0.50 (0.20) | 0.69 (0.19) | .01a | |
Total cholesterol (mg/dL), mean (SD) | 141.7 (22.1) | 180.2 (22.8) | <.001a | |
Total protein (g/dL), mean (SD) | 7.1 (0.5) | 7.2 (0.4) | .37a | |
Albumin (g/dL), mean (SD) | 4.3 (0.3) | 4.4 (0.3) | .17a | |
Total bilirubin (mg/dL), mean (SD) | 0.69 (0.30) | 0.65 (0.26) | .69a | |
Aspartate aminotransferase (mg/dL), mean (SD) | 26.4 (10.4) | 18.3 (3.1) | .001a | |
Alanine aminotransferase (mg/dL), mean (SD) | 33.1 (22.0) | 15.1 (7.9) | .02a | |
White blood count (no/mm3), mean (SD) | 5347.3 (1980.1) | 5857.0 (1397.6) | .46a | |
Hemoglobin (mg/dL), mean (SD) | 14.0 (1.44) | 13.6 (1.4) | .42a | |
Platelet (no/mm3), mean (SD) | 182,200 (118,600) | 276,600 (59,100) | .02a |
aDerived from Student
bDerived from Fisher exact probability test.
cbpm: beats per minute.
dDerived from Mann Whitney
eIQR: interquartile range.
Change of thyroid function and associating parameters during the study period in thyrotoxicosis and control groups.
Thyroid function test results and associated parameters | Visit 1 | Visit 2 | Visit 3 | Visit 4 | Visit 5 | |||||||
28 | 28 | 28 | 23 | 5 | — | |||||||
Free thyroxine (ng/dL), mean (SD) | 3.08 (1.09) | 2.02 (0.61) | 1.66 (0.64) | 1.60 (0.59) | 1.96 (0.39) | <.001b | ||||||
Thyroid stimulating hormone (mIU/L), median (IQRc) | 0.01 (0.00) | 0.01 (0.00) | 0.01 (0.13) | 0.01 (0.10) | 0.01 (0.10) | .114d | ||||||
Hyperthyroid Symptom Scale, median (IQR) | 12.5 (10.0) | 5.5 (7.8) | 4.0 (8.8) | 3.5 (4.5) | 7.0 (4.5) | <.001d | ||||||
On-site resting heart rate (bpme), mean (SD) | 101.6 (14.5) | 94.4 (15.4) | 90.1 (17.6) | 85.5 (12.0) | 83.5 (4.0) | .015b | ||||||
WD-rHR-ownf (bpm), mean (SD) | 88.0 (11.5) | 82.9 (10.9) | 75.9 (8.8) | 76.2 (8.0) | 72.7 (7.2) | <.001b | ||||||
WD-rHR-Fitbitg (bpm), mean (SD) | 82.2 (12.5) | 76.8 (9.5) | 70.8 (8.4) | 71.4 (7.8) | 74.7 (9.8) | <.001b | ||||||
10 | 10 | 10 | — | — | — | |||||||
Free thyroxine (ng/dL), mean (SD) | 1.36 (0.12) | 1.37 (0.13) | 1.35 (0.10) | — | — | .285b | ||||||
Thyroid stimulating hormone (mIU/L), median (IQR) | 1.33 (1.19) | 1.62 (2.21) | 2.06 (1.21) | — | — | .236d | ||||||
Hyperthyroid Symptom Scale, median (IQR) | 0.5 (2.8) | 0.0 (2.0) | 0.0 (2.0) | — | — | .504d | ||||||
On-site resting heart rate (bpm), mean (SD) | 81.9 (14.8) | 81.0 (13.2) | 76.4 (11.2) | — | — | .250b | ||||||
WD-rHR-own (bpm), mean (SD) | 65.8 (8.0) | 63.2 (7.4) | 64.5 (9.2) | — | — | .374b | ||||||
WD-rHR-Fitbit (bpm), mean (SD) | 66.5 (8.2) | 64.3 (7.3) | 64.3 (7.7) | — | — | .101b |
aVisit 1 to 3 were compared both in thyrotoxicosis and control groups.
bDerived from ANOVA with repeated measures with a Greenhouse-Geisser correction.
cIQR: interquartile range.
dDerived from Friedman test.
ebpm: beats per minute.
fWD-rHR-own: the resting heart rate from wearable device derived by own algorithm.
gWD-rHR-Fitbit: the resting heart rate from wearable device derived by Fitbit algorithm.
Linear model GEE analyses were performed to verify the relationship between serum free T4 levels and each associated parameter (ie, HSS, on-site rHR, WD-rHR-own, and WD-rHR-Fitbit). In these analyses, the mean (SD) of each parameter were standardized to 0 and 1, respectively, to compare each parameter’s relationship with free T4 levels and with each other. Before standardization, the 1SD of HSS, on-site rHR, WD-rHR-own, and WD-rHR-Fitbit were 6.3, 15.8, 11.4, and 11.2, respectively in thyrotoxic patients and 6.3, 16.2, 11.4, and 11.4 in all study participants. Although all parameters analyzed were significantly associated with serum free T4 levels in thyrotoxicosis patients and in all study participants, unstandardized beta for serum free T4 level by an increase of 1 SD of on-site rHR was relatively lower than those of other parameters (
The relationship between thyrotoxicosis, defined as 1.8 ng/dL or more of free T4, and each associated parameter was assessed with binary logistic model GEE analyses. All parameters analyzed were significant for predicting thyrotoxicosis (
Change of serum free thyroxine (T4) levels (A), hyperthyroid symptom scale (B), on-site heart rate (C), and resting heart rate from wearable device (D) during the study period. HSS: Hyperthyroid Symptom Scale, rHR: resting heart rate, WD-rHR-own: rHR from wearable device derived by own algorithm, WD-rHR-Fitbit: derived by Fitbit algorithm. Error bars represent 95% CI of the means.
Linear model generalized estimating equations analyses for the association between free thyroxine and associating parameters. Parameters standardized to the same mean and SD (mean 0, SD 1.0) for comparison and analyzed separately.
Association between free thyroxine and parameters | Unstandardized beta | 95% CI | ||
Hyperthyroid Symptom Scale | .504 | 0.333-0.674 | <.001 | |
On-site resting heart rate | .362 | 0.122-0.602 | .003 | |
WD-rHR-owna | .465 | 0.300-0.630 | <.001 | |
WD-rHR-Fitbitb | .513 | 0.331-0.694 | <.001 | |
Hyperthyroid Symptom Scale | .541 | 0.394-0.687 | <.001 | |
On-site resting heart rate | .396 | 0.204-0.588 | <.001 | |
WD-rHR-own | .492 | 0.367-0.616 | <.001 | |
WD-rHR-Fitbit | .515 | 0.375-0.656 | <.001 |
aWD-rHR-own: resting heart rate from wearable device derived by own algorithm.
bWD-rHR-Fitbit: resting heart rate from wearable device derived by Fitbit algorithm.
Binary logistic model generalized estimating equations analyses for the association between thyrotoxicosis and associating parameters. Parameters are standardized to have same mean and SD (mean 0, SD 1.0) for comparison and analyzed separately.
Association between free thyroxine and parameters | Odds ratio | 95% CI | ||
Hyperthyroid Symptom Scale | 2.614 | 1.612-4.238 | <.001 | |
On-site resting heart rate | 1.648 | 1.047-2.594 | .031 | |
WD-rHR-owna | 2.475 | 1.470-4.166 | <.001 | |
WD-rHR-Fitbitb | 2.863 | 1.649-4.970 | <.001 | |
Hyperthyroid Symptom Scale | 3.601 | 2.190-5.923 | <.001 | |
On-site resting heart rate | 2.114 | 1.365-3.273 | <.001 | |
WD-rHR-own | 3.840 | 2.113-6.978 | <.001 | |
WD-rHR-Fitbit | 3.843 | 2.067-7.144 | <.001 |
aWD-rHR-own: resting heart rate from wearable device derived by own algorithm.
bWD-rHR-Fitbit: resting heart rate from wearable device derived by Fitbit algorithm.
To evaluate the clinical feasibility of wearable device generated data for the management of thyrotoxicosis, we investigated the association between HR data collected by activity trackers and thyroid function across the clinical course of thyrotoxicosis. Our results demonstrated that both rHR measured by a wearable device and HSS were significantly associated with thyroid function.
It is well known that the thyroid hormone has a positive chronotropic, inotropic effect, meaning that it stimulates the rate and force of systolic contraction and the rate of diastolic relaxation [
HSS, on-site rHR, WD-rHR-own, and WD-rHR-Fitbit were decreased as thyrotoxicosis improved and these variables were all associated with serum free T4 levels and thyrotoxicosis in linear and binary logistic model GEE analyses, respectively. Interestingly, standardized on-site rHR showed relatively lower beta for serum free T4 level and a lower OR for thyrotoxicosis compared with other standardized parameters in the GEE analyses. On-site rHR is a one-time measurement of one clinical aspect, while HSS is a one-time measurement of several clinical aspects and WD-rHR is calculated from continuously monitored data for one clinical aspect. These differences can explain on-site rHR’s relatively weak correlation with thyroid function compared with HSS or WD-rHR. Similar beta and ORs and their 95% CIs between standardized HSS and WD-rHR suggest that WD-rHR calculated from continuously collected, detailed data by wearable devices may be used to assess thyroid function status with similar accuracy and precision with HSS, a validated assessment tool known to reflect thyroid function, but which also requires time and an educated scorer [
Current wearable devices allow highly detailed, longitudinal measurement of physical indices. This large volume of clinical information (ie, “high definition data”) may provide more accurate and objective information than subjective symptoms or physical signs recorded at clinic visits, which can be influenced by diurnal variation, emotional state, and other factors [
This study has some significant clinical implications. Although monitoring thyroid function using wearable devices cannot replace TFT, it could aid the management of thyrotoxicosis. During the treatment of GD, most patients are expected to repeat the TFT every one to two months regardless of their response to ATDs, and the interventions including dose adjustment of ATDs are provided with the same time interval. Monitoring thyroid function using wearable devices may provide patients with individualized, flexible, and more accurate interventions during treatment and follow-up, minimizing inconvenience and costs. Moreover, about 50% of patients have been reported to relapse within two years after discontinuing ATD, even if they were treated according to the recommended guidelines [
The primary strength of the study was that it contributes to the literature by monitoring HR throughout the day using wearable devices, for the first time, in patients with thyrotoxicosis. The clinical evidence provided herein may inspire further investigations and clinical applications of biosignals monitored by wearable devices in thyroid dysfunction. The other strength is that we used commercially available wearable devices and mobile applications. Thus, our results can be immediately applied to the management of thyrotoxic patients and provide a ready-to-use algorithm for making Web-based or mobile applications for managing thyrotoxicosis.
This study also has some limitations, which should be considered when interpreting the results. First, it was likely easier for younger people, who are adept at using mobile phones and wearable devices, to participate in the study, although the participants’ ranged from 18 to 60 years of age. Also, signs such as increased HR owing to thyrotoxicosis may not be as evident in the elderly [
In conclusion, our results indicate that rHR data from wearable devices show reasonable predictability of thyroid function in patients with thyrotoxicosis. This parameter can be measured relatively simply and may be useful as well as clinically feasible for the management of thyroid dysfunction. This study is a starting point for the clinical application of high-definition medicine in the management of thyroid disease.
Inclusion and exclusion criteria, biochemical measurements, and generalized estimating equations analyses.
analysis of variance
antithyroid drug
case report form
differentiated thyroid cancer
Graves’ disease
generalized estimating equation
heart rate
Hyperthyroidism Symptom Scale
interquartile range
JavaScript object notation
odds ratio
Seoul National University Bundang Hospital
triiodothyronine
thyroxine
thyroid function test
thyroid hormone stimulating hormone
resting heart rate from wearable device derived by Fitbit algorithm
resting heart rate from wearable device derived by own algorithm
Funding for this study was provided by Seoul National University Bundang Hospital, which had no role in designing the study, data collection, analysis, or interpretation, writing the report, or decisions regarding submitting the paper for publication. This research was also supported by a research grant (02-2016-035) from the Seoul National University Bundang Hospital to JHM.
JHM was responsible for conception and design of the work. JHM, JEL, KMK, and TJO collected the data. JHM and JEL analyzed the data and drafted the manuscript. JHM, JEL, DHL, SHC, SL, YJP, DJP, and HCJ contributed data interpretation. JHM and JEL approved the final version of the manuscript. All authors agreed on the final content of the manuscript.
None declared.