Re: “Association Between Primary Hypothyroidism and Nonalcoholic Fatty Liver Disease: A Systematic Review and Meta-Analysis” by Mantovani et al . ( Thyroid 2018;28:1270–1284)

Dear Editor:

We read with interest the manuscript by Mantovani et al. assessing the association between hypothyroidism and nonalcoholic fatty liver disease (NAFLD) (1). In this meta-analysis, including 12 cross-sectional and three longitudinal studies, the authors showed that hypothyroidism was significantly associated with NAFLD (odds ratio [OR] = 1.42 [confidence interval (CI) 1.15–1.77]), as well as with the presence of nonalcoholic steatohepatitis or advanced fibrosis (OR = 2.73 [CI 1.90–3.93]).

While the authors should be congratulated for a well performed study, an important issue deserves further consideration. In the subset of studies included in this meta-analysis, hypothyroidism was defined in very different ways, including overt hypothyroidism, subclinical hypothyroidism, and/or levothyroxine replacement. The authors did a good job providing a sensitivity analysis that divided the studies based on the definition of hypothyroidism used. However, results from these analyses only unveil new questions.

If low free thyroxine (fT4) or high thyrotropin (TSH) levels (i.e., subclinical or overt hypothyroidism) are responsible for the changes observed in hepatic steatosis and liver histology, how do the authors explain that patients receiving levothyroxine replacement actually showed a higher risk of NAFLD than those not receiving treatment (OR = 2.19 [CI 1.41–3.43] vs. OR = 1.31 [CI 1.04–1.66])? If we assume that most of the people defined as hypothyroid based on levothyroxine replacement were actually euthyroid, what are potential mechanisms that explain their higher prevalence of NAFLD? Potential explanations for this finding seem rather complicated.

In order to avoid this issue of combining patients with untreated subclinical/overt hypothyroidism and those receiving treatment, in our cohort of 232 patients with type 2 diabetes mellitus, we excluded patients receiving levothyroxine replacement (2). In that scenario, we observed that even when TSH levels were normal, low fT4 levels were associated with increasing intrahepatic triglyceride content as assessed by proton magnetic resonance spectroscopy, a much more accurate way of measuring liver fat than ultrasound. However, we observed no significant associations between fT4 or TSH and liver histological parameters, such as inflammation, hepatocyte ballooning, or fibrosis (2).

Observational studies such as the ones included in this meta-analysis are prone to several confounding factors that cannot be perfectly accounted for only with statistical adjustments. In support of this, when longitudinal studies were assessed in the meta-analysis, no association between hypothyroidism and the development of NAFLD was observed. Therefore, results from this study should be taken with caution. However, we agree with the overall message of the authors, suggesting that strong consideration for NAFLD should be given to patients with hypothyroidism, especially if overweight or obese.