We read with interest the work of González et al. (1), which to our knowledge is the largest study of the predictive value of thyroid antibody testing for future thyroid dysfunction in patients with type 1 diabetes. We previously examined this in a retrospective study of 467 patients with diabetes (81 with type 1 diabetes) and found that, although the presence of thyroid autoantibodies at diagnosis predicted later thyroid dysfunction, a high-normal baseline thyroid stimulating hormone (TSH) level (i.e., above the 75th percentile) was more predictive (2). When both were combined in the same statistical model, only the baseline TSH level remained a statistically significant predictor. We speculated that this was because a high-normal TSH represents early thyroid disease and that the TSH assay may have been more sensitive than the antibody assays.
It would be interesting to know whether a similar pattern was seen in the data of González et al.
