2016 年 4 巻 1 号 p. 32-38
Background: Patients with thyroid dysfunction (PTDs) are likely to go undiagnosed. Therefore, we developed a new, low-cost screening method using a set of six routine tests to identify PTDs. This new method can predict the probability of thyroid dysfunction at the time of the screening. We found a lot of new PTDs through our screening method in general health check-ups and determined that an 85% threshold for the predicted probability was sufficient, instead of the 60% probability used in the screening. However, there were about three times the number of false positives compared to the number of true positives, mainly due to the effects of individual differences and background diseases.
Objective: The aim of the present study was to develop a method considering time-series variations in routine tests in addition to the predicted probability, which had only been used previously, to improve predictive accuracy.
Methods: The present study included 155 subjects (15 true positives and 140 false positives) who were suspected of having thyrotoxicosis in our screening, whose thyroid hormones were measured, and they also had previous visit records. We calculated the average rate of annual time-series variations (RATV) in each routine test between previous and current visits for each subject, and then plotted RATV and predicted probabilities together in a scattergram.
Results: By adjusting both thresholds on the scattergram, the following optimized thresholds were obtained: 98% for predicted probability or 20% for RATV. This combination significantly decreased the number of false positives from 140 to eight without yielding any false negatives for Graves’ disease.
Conclusions: The introduction of RATV markedly reduces the influence of individual differences and background diseases in routine tests.
