Subclinical hypothyroidism is associated with migraine: A case-control study;methodological and statistical issues

To the Editor:

We read with interest the study by Elisa Rubino et al. (1) entitled “Subclinical hypothyroidism is associated with migraine: A case-control study” published in Cephalalgia. The authors concluded that “migraine is more frequent in patients with subclinical hypothyroidism in respect to controls”. However, there are several methodological and statistical issues threatening the validity of the reported findings.

First, although the authors had measured several covariates including age, sex, body mass index (BMI), smoking, hypothyroidism-inducing medications, hypertension, and further auto-immune diseases, there are only unadjusted ORs in the results section. Due to the possible effect of confounding variables, it would be of great value for the readers to see also the adjusted ORs using a multiple logistic model.

Second, the type of control group in the abstract is not clear. This is important for readers in order to draw a better conclusion (2,3). There are some reported comparisons between subclinical hypothyroidism patients with and without migraine in the abstract. This led to an ambiguity regarding the nature of the study controls, that is, the relatives of patients.

Third, the response rate for case and control groups was not reported. Moreover, there is no information regarding the actual coverage of the “Endocrinological Clinic of the Umberto I Hospital of Torino”. Without this information, it would be hard to draw a correct conclusion regarding the possible selection bias of the study (2,3).

Fourth, the validity indices, that is, the sensitivity and specificity of the “Italian version of the Structured Migraine Interview (SMI)” as well as the “ICHD-III beta version criteria” have not been reported. Therefore, it would be again hard to draw conclusions about the possible information bias (3). It is highly recommended to clarify the possible biases (selection, information and confounding) in the limitations section of the study.