Abstract
We thank Lutz Schomburg for his expert opinion and useful comments on our meta-analysis regarding selenium (Se) supplementation in the treatment of Hashimoto thyroiditis (HT) (1).
We agree with his concerns regarding the limited number of existing randomized, placebo-controlled trials (RCTs). As stated in the discussion section, “limited primary data” may lead to an analysis “that could be underpowered” or “vulnerable to the distorting effect of a single biased estimate” and ultimately to “publication bias.” Furthermore, we concur with Schomburg's suggestion that nutritional Se supply might be a potential effect modifier. In fact, this is discussed by all primary investigators. As stated in the meta-analysis, however, Se levels were determined in only two of the included studies (2,3), and no correlation between Se levels and the response to treatment was documented in any of them. Although reasonable, Schomburg's assumption about the lack of effect in Se-sufficient populations needs to be confirmed in RCTs. Moreover, Schomburg questioned the need for a subgroup analysis on the basis of Se preparation (selenomethionine versus selenite). Given the differences in the pharmacodynamic profile (4) and metabolism between the two preparations, we believe that the two compounds cannot be regarded as identical and differences in efficacy cannot be precluded. In fact, his group recently provided evidence documenting compound-specific differences in toxicity among 10 different seleno-compounds in vitro (5).
Schomburg correctly emphasized that there is an upper limit for Se supplementation, covering an omission in the original publication. In support of the statement that in regions without Se deficiency, Se supplementation could produce toxic effects, he cited a prospective trial, with Se supplementation in nonmelanoma skin cancer patients in which incidence of type 2 diabetes mellitus (T2DM) was a secondary outcome (6), and a cross-sectional study investigating the association between baseline Se levels and T2DM incidence (7). Both studies suggested a slightly higher risk of T2DM in males with higher Se levels. Schomburg suggested that the higher incidence of T2DM was not noted in our meta-analysis because of the female predominance in HT and the relative Se deficiency and proposed a very interesting and useful outline for future research. We agree that this contribution is important. However, it should be noted that the cited prospective trial has several methodological issues, regarding lack of report on baseline fasting and postprandial blood glucose levels, hemoglobin A1c levels, and family history of diabetes that could have resulted in unbalanced treatment groups (
More importantly, our main disagreement hinges on Schomburg's interpretation of the meta-analysis findings. By stating “… no risks, just benefits?” and “therapeutic success,” it is implied that Se supplementation had beneficial effects in the course of HT. On the contrary, our meta-analysis demonstrated the “lack of outcomes with direct clinical implications,” such as thyroid function, thyroid ultrasound morphology, or demands in thyroxine replacement therapy, interpreting the significant decrease in thyroid peroxidase antibody levels as “an intermediate outcome of limited clinical importance in the course of HT ….” In our opinion, these findings should not be considered as therapeutic success. Having said that and endorsing Schomburg's comment on an upper limit for Se supplementation and future research, we believe that this communication greatly contributed to improving the meta-analysis message.