Discussion
Pregnant women in mid-pregnancy period are prone to ID2. Serum ferritin plays an important role in reflecting the iron reserves4. Our study showed significant correlations between thyroid hormones and iron status. Although there were some correlations between thyroid hormones and iron status factors, there were not completely consistent in a previous study. Shan et al. pointed out that ID women had a lower FT4 levels, but had no significant difference in TSH levels19. We did not have evidence that directly supported the specific relationship between thyroid hormones and iron nutrition. However, our results and those of previous studies indicated that close relationship between thyroid hormone and iron nutrition. Especially, our results can be considered as a certain reference due to be based on a large sample size.
In a previous study that established the trimester-specific reference interval of thyroid hormones in pregnant women, it was pointed out that the FT4 RI in the second trimester was 11.70 (8.67, 16.21) pmol / L, and the TSH RI in the second trimester was 1.64 (0.27, 4.53) μIU/mL20. The RIs of thyroid hormones in our study differed from those in previous reports because of differences in detection reagents, races, detection methods or kits,and individual differences. Compared with other studies, the reference critical values mentioned in our study were slighter higher. However, the IRs of each parameter established in this study had certain reference significance.
In recent years, the number of studies on the relationship between thyroid and iron nutrition increased. Some papers pointed out that the high level of TSH and the low level of FT4 were independently associated with ID in pregnant women21. Other studies also showed that SF was an important predicter of TSH during the second trimester of pregnancy4. ID might result in higher TSH and lower FT4, FT3 in the pregnancy, which affects the thyroid function of pregnant woman22. Some researchers found that ID decreases the level of SF, FT3, FT4, and enzyme activity of TPO in rats23. However, a study indicated no significant correlation was found between thyroid function and ferritin in pregnant women from Indonesia24. And a study pointed out that iron nutrition was positively correlated with FT4 during pregnancy, except in the third trimester25. In the current study, we observed that women in ID group who were assessed for SF levels during the second trimester pregnancy showed lower levels of FT3 and FT4 and a higher level of TSH. Our results were consistent with those of most previous studies. Therefore, comparison of the findings with other previous studies could confirm that the iron nutrition status was positively associated with thyroid hormones during the second trimester pregnancy. However, it is unclear how iron status interacts with thyroid functions specifically. Thus, further research is needed.
Although iron status is not the only factor affecting thyroid function, it has a significant auxiliary effect4. Several studies indicated that ID influenced thyroid function by the following mechanism. Some studies pointed out that the decline in FT4 concentrations might be associated with the change of the TPO activity; TPO is an iron-dependent enzyme that plays an important role in the production of thyroid hormones; therefore, ID might cause lower FT4 levels via reducing TPO activity; then, the levels of TSH increase in negative feedback26-28. Briefly, the lower FT4 levels secondary to ID results in the decrease in FT4 negative feedback inhibition of TSH levels, thereby increasing TSH levels16.
When ID develops to some extent, it will result in ID anemia. Thus, as an anemia index, Hb is also involved in this study to compare the influence on thyroid function with SF (Hb < 110g/L were found in 59% in ID group). Interestingly, Spearman correlation analysis showed that both SF and Hb positively correlated with FT4 and FT3 and negatively correlated with TSH. Yet canonical correlation analysis reflected the relationship between SF and FT3. Therefore, we were sure that iron nutritional status was associated with thyroid hormone, and we predicted SF and Hb were risk factors that contribute to thyroid hormones. Furthermore, as the results of canonical correlation contradicted with the results of regression, whether SF or Hb plays a key role in affecting the thyroid hormones is uncertain if our data analysis result is the only basis. Previous experiments suggested that Hb synthesis occurs prior to TPO when ID occurs29. Thus, we also speculated that when ID occurs, iron will be preferentially supplied to the maternal hemoglobin to maintain current levels, which results in lower SF levels and decreases synthesis and activity of TPO. Then, thyroxine insufficiency arises, especially the FT4 and FT3 levels. Lower FT3 and FT4 levels are obtained, thereby raising TSH levels via negative feedback inhibition.
Canonical correlation analysis is a useful statistical method to analyze the interrelationships between subsets of multiple dependent and independent variable quantities30. It could be used to analyze the complex interactions of date from two subsets of variable quantities31. In our study, to reduce the influence of collinearity and highlight the main relationship in two sets, canonical correction analysis was added. Considering that the units of variables studied were different, the standardized correlation coefficients were used. The larger the standardization coefficient was, the greater the weight reflected in typical variables in its subsets. The fact that both the standardized and the structure coefficients of SF and FT3 were the highest in their respective subsets reflected the relationship between SF and thyroid functions. SF was positively correlated with FT3 and FT4 and negatively correlated with TSH. In addition, the results of standardization coefficient and structure coefficient were consistent in our study.
The elastic network model is one of the most important models on nonlinear prediction. This model is controlled by two parameters, namely, Alpha and Lambda32. Being different from ridge and lasso regression, the elastic network model merges both regularization skills within the one model; thus, it not merely lessens the feature coefficients but also installs some of the coefficients to zero to reduce the dimensionality of the feature space; then, the potentially meaningful correlation structure also could be preserved32, 33. In our study, none of the included variables were excluded, and the coefficient of SF is slightly smaller than that of Hb. The results of elastic network contradicted with the results of canonical correlation, maybe because both SF and Hb played certain roles in regulating thyroid function. However, they are all indicators of iron nutrition. Thus, iron nutritional status is associated with thyroid hormone levels in the second trimester of pregnancy.
The innovations in our study were pointed out as follows. First of all, our result was based on larger samples from the same regions, and all of recruited women were in the second trimester of pregnancy. Throughout the whole pregnancy, the thyroxin level in serum is fluctuating. For example, due to the change of human chorionic gonadotropin, the level of FT4 elevated temporarily and then declined gradually34. To our knowledge, this is the first study to compare the influence of SF and Hb on thyroid hormones via canonical correlation and elastic network regression analysis. Our study had some limitations. Firstly, our study was based on the results of large population survey. Thus, there was a lack of experimental for further demonstrations. However, the change tendency of our results was consistent with previous studies and experiments. Then, we did not measure the levels of urine iodine or thyroid antibodies. Thus, we could not perform a comprehensive evaluation of pregnant women. Nevertheless, it will be beneficial to perform similar studies in different trimesters of pregnancy to do further study the relationship between iron nutritional status and thyroid function.