The optimal dosing scheme for levothyroxine after thyroidectomy: A comprehensive comparison and evaluation

Area of Science:

• Endocrinology
• Pharmacology
• Medical Informatics

Background:

• Achieving euthyroidism after thyroidectomy can be challenging for patients.
• Standard weight-based levothyroxine dosing may not be optimal for all individuals.
• Novel dosing schemes are being explored to improve initial levothyroxine prescription accuracy.

Purpose of the Study:

• To review existing levothyroxine dosing schemes for patients post-thyroidectomy.
• To develop and compare novel machine learning-based dosing schemes.
• To evaluate the accuracy of different schemes in predicting the euthyroid levothyroxine dose.

Main Methods:

• Retrospective analysis of 598 patients who achieved euthyroidism after thyroidectomy.
• Scoping review to identify existing levothyroxine dosing schemes.
• Application of 13 machine learning algorithms, including Poisson regression, for dose estimation.
• 10-fold cross-validation to compare prediction accuracy against actual euthyroid doses.

Main Results:

• A novel Poisson regression model demonstrated the highest accuracy, predicting the correct dose for 64.8% of patients.
• The Poisson regression scheme significantly outperformed the best literature-based scheme (60.9% accuracy, P=.031).
• Standard weight-based dosing (1.6 µg/kg/day) achieved 51.3% accuracy, while an age/sex/weight-based scheme predicted only 27.4%.

Conclusions:

• A new Poisson regression dosing scheme, utilizing readily available variables, is more accurate than existing methods and other machine learning algorithms.
• This machine learning approach offers a superior method for estimating initial levothyroxine dosage after thyroidectomy.
• The findings suggest a potential improvement in achieving and maintaining euthyroidism in patients following thyroidectomy.