SMART: Sensor data and machine learning to aid hormone replacement of the thyroid
Description
Thyroid hormones regulate energy metabolism in nearly all tissues of the body. While about 5% of the population has clinical hypothyroidism, an additional 5% has subclinical disease. The treatment of this chronic disease is to replace the thyroid’s function with synthetic hormones. Five percent of the population is using thyroid hormones due to thyroid dysfunction or previous surgery, making levothyroxine (LT4) one of the three most prescribed medications in Europe (including Norway) and North America.
In this project, we will study how the use of hormone replacement drugs can be optimized to improve the quality of life and ability to work. We will collect data in a clinical trial, but also use existing Norwegian health care data to evaluate patients’ health outcomes and economic potential of treatment improvements. As the loss of thyroid function is not reversable, suboptimal treatment can have several long-term consequences, including cardiovascular disease, dementia, osteoporosis and psychic unhealth. The proposed project is highly ambitious because it seeks to renew a longstanding medical practice. We will conduct a clinical study that includes 240 patients in nearly all ages of adulthood (18-80 y). By building on a previously developed clinical decision support system, we will join clinicians, machine learning experts and innovative technology to improve the quality of treatment for a large patient group that consumes considerable health care resources, especially in primary health care.
Goals
Primary objective:
- To develop more accurate thyroid hormone replacement therapy.
Secondary objectives:
- To investigate if noninvasive sensor data can help predict the optimal dosage of thyroid replacement hormones.
- To investigate if noninvasive sensor data can be used as an objective measure of successful thyroid hormone replacement therapy.
- To measure how treatment optimization affects quality of life and societal benefits.
- To develop a decision support tool to suggest adjustments in the thyroid replacement hormone dosage.
Method
To achieve these objectives, in SMART we propose a new method to objectively assess if hormone levels are properly restored and optimize thyroid hormone replacement therapy. Concretely, this project will exploit the information obtained from noninvasive sensors recording electrocardiogram (ECG), since the heart rate variability (HRV) decreases during hypothyroidism and increases during hyperthyroidism. Over the last decade, machine learning algorithms have been established as a powerful tool to analyze and classify ECG signals. We will use machine learning techniques to predict the optimal dosage of thyroid replacement hormones from noninvasive sensor data.