Construction of a Nomogram Model to Predict The Risk of Hypothyroidism in Patients With Thyroid Nodule Caused by Environmental Iodine and Fluorine


 Background: Patients with thyroid nodules may be complicated by hypothyroidism. However, effective means to predict thyroid-stimulating hormone (TSH) elevation in patients with thyroid nodules are currently lacking. Therefore, the research team recruited some patients with thyroid nodules to explore the relationship between TSH, environmental iodine, and fluoride and the development of thyroid nodules, and to draw a Nomogram in order to provide guidance for predicting hypothyroidism in patients with thyroid nodules.Methods: The subjects were from 313 patients with thyroid nodules in different areas of iodine and fluoride, with 71 men and 242 women. The content of TSH in serum of thyroid nodule patients was determined by electrochemical luminescence method. The iodine content in drinking water and urine was determined by arsenic-cerium catalytic spectrophotometry. Fluoride in drinking water and urine was determined by fluoride ion selective electrode method. Univariate analysis, Lasso regression analysis and multiple factor Logistic regression analysis were used to screen the variables included in the Nomogram model. And the corresponding Nomogram model was constructed. ROC curve was drawn and repeated Bootstrap self-sampling method was used to test model differentiation.Results: There were 53 patients with elevated serum TSH and 256 patients with normal thyroid nodule. Multivariate logistic regression analysis shows that urine iodine（Odds Ratio[OR]=1.001, 95% confidence interval [95%CI]:1.001～1.002, P=0.01） and gender (OR=3.328, 95% CI:1.256～8.819, P=0.079) are independent risk factors raised by serum TSH. And drinking water fluoride（OR=1.990, 95% CI:0.924～4.289, P=0.079）is a critical factor. Based on these risk factors, the predictive model and nomogram of serum TSH elevation in patients with thyroid nodule were constructed. The area under the ROC curve was 0.678 (P<0.001), confirming Nomogram's predictive potential. The calibration diagram of the Nomogram prediction model shows that the predicted values are in good agreement with the observed values.Conclusion: Gender, urine iodine and drinking water fluoride of thyroid nodules are independent influencing factors for the rise of serum TSH. The Nomogram can become a predictive accuracy and differentiation.


Introduction
Thyroid nodule is one of the most common thyroid diseases and it is a localized mass with one or more abnormal tissue structures caused by various reasons. In recent years, the prevalence has increased rapidly [1,2] [3]. The etiology is complicated and related to gender, age [4], iodine [5] and uorine [6,7] intake, history of radiation exposure [8], and immune factors [9]. High or low iodine intake and high uoride intake can cause thyroid nodules. At the same time, long-term stimulation of thyroid-stimulating hormone (TSH) may promote thyroid hyperplasia and form benign and malignant nodules. Studies have found that detecting serum TSH in patients with thyroid nodules is of great signi cance in assessing the properties of thyroid nodules [10,11] Thyroid nodules can be complicated by hypothyroidism [3]. Hypothyroidism is a metabolic abnormality caused by reduced metabolism or insu cient secretion of thyroid hormones. It includes subclinical hypothyroidism (referred to as subclinical hypothyroidism) and thyroid. Hypofunction (abbreviated hypothyroidism). The survey shows [3] that the incidence of subclinical hypothyroidism is 16.7%, which is increasing year by year. Subclinical hypothyroidism is a high-risk factor for hypothyroidism And the incidence of subclinical hypothyroidism is increasing year by year. The prevalence of hypothyroidism among residents has also increased.
TSH is secreted by the pituitary gland, which can promote the proliferation of thyroid follicular epithelial cells, leading to thyroid hyperplasia, which is the main indicator for evaluating thyroid function. Elevated serum TSH levels can induce the occurrence of thyroid nodules and increase the incidence of thyroid cancer. Elevated serum TSH levels may also increase the risk of hypothyroidism [12].
Although there have been related reports, the relationship between thyroid nodules, iodine and uorine nutrition status, and the change rule of TSH in thyroid nodule patients are not clear; The results of studies on the effects of iodine and uorine intake on hypothyroidism are inconsistent [7,13,14] [15][16][17][18][19][20][21] . There is also a lack of an effective means to predict hypothyroidism based on changes in serum TSH. In view of this, we enrolled some patients with thyroid nodules in different water-derived iodine and uorine regions to evaluate their iodine and uorine nutrition status. The serum TSH level was detected to explore the relationship between TSH, environmental iodine and uorine and the incidence of thyroid nodules. A Nomogram was drawn in order to provide a basis for the prevention of thyroid nodules in patients with thyroid nodules.
The ultrasound doctor used the color Doppler ultrasound of the Mindray Z5 and 75L 38EA probes to complete the thyroid examination .And according to the "China Guidelines for thyroid nodules and differentiated thyroid cancer (2012) [23], urine and blood samples were collected and tested in patients diagnosed with thyroid nodule by color ultrasound and those without obvious clinical symptoms. Inclusion criteria were as follow: 1) the age range was 25 to 80 years old. 2) the residents who were born and have lived continuously in the country for 25 years or more. 3) Thyroid nodule was diagnosed by color Doppler ultrasonography and serum thyroid stimulating hormone (TSH) was detected for the rst time. Exclusion criteria were as follow: the people who were taking iodine-containing drugs, suffering from thyroid cancer, thyroid surgery and person with serious diseases, or have incomplete urine or blood samples. or pregnant women. A total of 313 thyroid nodule patients, 71 male and 242 female, were recruited from seven administrative villages with different levels of iodine and uoride in drinking water.

Collection and inspection of water samples
Each administrative village where the volunteers were located had centralized water supply. Two terminal water samples were collected from each administrative village and the iodine uoride content in the water samples was determined. The mean value of the detection results of the two water samples, namely, the iodine uoride content in drinking water of the administrative village was determined by arsenic cerium catalytic spectrophotometry (GB/T5750.  (http://down.foodmate.net/standard/sort/3/11211.html).The iodine content in drinking water was determined by uoride ion selective electrode method (GB/T5750.  (http://down.foodmate.net/standard/sort/3/11211.html).

Blood sample collection and inspection
Fasting venous blood of the subjects was collected in the morning. And the serum, TSH content was determined by electrochemical luminescence method using Roche CoBAS-E-602 electrochemical luminescence instrument (Roche Diagnostics Germany). Serum TSH kit is provided by Roche Diagnostics (Shanghai) Co. LTD. The test method is operated according to the instructions. Subjects were diagnosed as TSH increases [24], whose serum TSH is higher than 4.2uIU/ Ml (Serum TSH reference range : 0.27 ~ 4.2uIU/ mL).

Statistical analysis
The measurement data should be tested for normality rst. The Student's t-test was applied to text continuous and normal variables,and the results were expressed as the mean ± standard deviation (SD). Skewed distribution data are given by M ( , ), using Mann-Whitney U Test. Categorical variables are represented by number of cases or percentages, using two independent samples chi-squared test or Fisher's exact probability method. Univariate analysis was performed on the collected predictors to assess the correlation with hypothyroidism. Predictors were included in the Lasso regression analysis to screen for important risk factors for hypothyroidism. Predictions of P ≤ 0.05 in the univariate logistic regression analysis were included. Variables were combined with predictive variables with non-zero coe cients in the Lasso regression analysis. And multivariate logistic regression analysis was performed to screen independent risk factors affecting hypothyroidism. Based on the results of multivariate logistic regression analysis, a Nomogram model was established. Bootstrap resampling the Nomogram model is used to verify the prediction performance. The consistency coe cient (C-index) is used to represent the prediction performance of the Nomogram model and a calibration chart is drawn to indicate its prediction compliance. The receiver operating characteristic curve (ROC curve) is plotted to obtain the area under the ROC curve (area under the cure (AUC) and its 95% CI, evaluating the prediction model.
Use R software to draw a line chart. In the application process, you can nd the corresponding score value according to each predictor of the patient in the line chart. Adding the scores of each predictor variable is the total score, which corresponds to the total score.

The characteristics of the patients with thyroid nodules
The characteristics of the selected population are shown in Table 1. Among 313 patients with thyroid nodules, 57 patients with abnormal serum TSH were found by serum TSH test (53 patients with elevated serum TSH and 4 patients with reduced TSH) and 256 patients with serum TSH within the reference value range (Table 1). Urinary iodine and uoride: The morning urine.
Drinking water iodine and uorine: Peripheral water.

Univariate analysis
Univariate analysis was performed using serum TSH in patients with thyroid nodules as predicted variables, and six factors as age, gender, urinary iodine, urinary uoride, drinking iodine and drinking water uoride as predictors. The results showed that urine iodine and urine Fluorine and gender may be the in uencing factors of serum TSH elevation in patients with thyroid nodules (P < 0.05), and iodine drinking water as the in uencing factor is critically signi cant (P = 0.061). (See Table 2) Through Lasso regression analysis, the relatively unimportant independent variable coe cients become 0 to exclude from modeling; those with non-zero variable coe cients are included as the most important risk factors in the modeling. The dependent variable was whether the serum TSH in patients with thyroid nodules was increased, and the independent variables were age, gender, urine iodine, urine uoride, iodine in drinking water and uorine in drinking water. LASSO regression analysis was performed. The results showed that the coe cients of the two variables, age and water iodine, became 0. There were four variables with non-zero coe cient, namely, gender (coe cient = 0.857), urine iodine (coe cient = 0.0008), urine uorine (coe cient = 0.050) and drinking water uorine (coe cient = 0.470). ( See Table 3, Fig. 2 and Fig. 3) In Lasso regression analysis, if the variable coe cient becomes zero, it will be excluded from modeling; If the variable coe cient is non-zero, it is included into the modeling as the most important risk factor.

Multivariate logistic regression analysis
Based on the results of univariate analysis and lasso regression analysis, ve factors including urinary iodine, urinary uoride, gender, drinking water uoride and drinking iodine were included in multivariate logistic regression analysis. The results showed that gender and urine iodine increased serum TSH in patients with thyroid nodules Independent in uencing factors, drinking water uoride was the critical in uencing factor. The risk of serum TSH in women with thyroid nodules is 3.328 times higher than that in men (95% CI:1.256 ~ 8.819, P < 0.05).In patients with thyroid nodule, the risk of serum TSH increased by 1.001 times for each increase of urine iodine concentration unit (95%CI:1.001 ~ 1.002, P < 0.05).The increased risk of serum TSH in thyroid nodule patients with high uoride content in drinking water was 1.990 times that in normal drinking water (95%CI:0.924 ~ 4.289, P = 0.079). (See Table 4)  Table 4).
To more intuitively predict the risk of elevated serum TSH in patients with thyroid nodules, a Nomogram was drawn. The Nomogram includes: prediction index, score scale corresponding to single prediction index, total score scale, risk value line of serum TSH increase etc. Three predictors were included in this study: gender, urinary iodine, drinking water uoride. The three predicted indicators were compared with the corresponding score scale to obtain the score. Sum of the scores of the above indicators to obtain a total score. The total score corresponds to the score on the predicted risk value line of serum TSH increase, that is, the predicted risk value of serum TSH increase. (See Fig. 4, Table S5 and Table S6) 2

.6 Evaluation of Nomogram model prediction capabilities
The degree of discrimination of the model is evaluated by the consistency index (C-index). The initial Cindex in this study is 0.678 (95% CI: 0.604 ~ 0.753). The receiver operating characteristic curve (ROC curve) is drawn to obtain the ROC. The area under the curve (AUC) is 0.678, P < 0.001.And the ROC analysis con rmed Nomogram's predictive potential. To see Fig. 5, a calibration diagram of a Nomogram prediction model is shown. There is good agreement between Nomogram's prediction and actual observation,indicating that the Nomogram model can be compared with Accurately predict the risk of elevated serum TSH in patients with thyroid nodules. (See Fig. 6)

Discussion
Studies have shown that the prevalence of thyroid disease is on the rise [1][2][3]. Thyroid nodules are one of the common diseases of thyroid disease and patients do not have obvious clinical symptoms in the early stage. They are mostly accidentally found during physical examination. The detection rate of color Doppler ultrasound can reach 19% ~ 68%, of which 7% ~ 15% are thyroid cancer [25,26]. And it may also be accompanied by changes in thyroid function and thyroiditis. Our previous work found that the prevalence of thyroid nodules in people over 30 years of age in Cangzhou City, Hebei, China ranges from 25.1% to 43.3% [22]. The etiology of thyroid nodules is complex. And it is related to gender [4], age [4] , iodine nutrition level [4,5] , high uoride [16,20,27] as well as history of radiation exposure [8]. and immune factors [9] etc. But it is not fully explored at present. There are reports that environmental iodine and uorine may be one of the risk factors [20,25]. Most of the subjects in this study came from watersource areas with high iodine content and low iodine content and high iodine content and high uorine content. Through more Logistic regression analysis of factors found that gender and urinary iodine were independent factors in uencing serum TSH elevation in patients with thyroid nodules, and drinking water uorine was the critical factor. Therefore, the study of the etiology of thyroid nodules can help to explore the effects of hypothyroidism and thyroid cancer risk factors.
The Nomogram is based on multivariate analysis and integrates multiple forecasting indicators, which can transform complex formulas into intuitive graphs. As a reliable statistical model, the Nomogram has been widely used in industrial and atmospheric research. In the medical eld, researchers [28][29][30] have established a Nomogram of the survival probability of different cancers. Clinicians can predict the survival probability of cancer patients after surgery based on the Nomogram, reduce the occurrence of overtreatment, rationally use medical resources, and reduce patients' pain. The established Nomogram can effectively predict the overall survival of patients with hepatocellular carcinoma after hepatectomy [31]. But there is no report about the prediction model of serum TSH increase in patients with thyroid nodules. The Nomogram C-index established in this study was 0.678 (95% CI: 0.604 0.753). The ROC curve AUC of the Nomogram prediction model is 0.678, indicating that Nomogram has certain practical value for predicting the risk of serum elevation in patients with thyroid nodules. The calibration chart of Nomogram shows that using Nomogram to predict the risk of elevated serum TSH in patients with thyroid nodules is more accurate. In clinical work, it is also expected that it can provide clinicians with an intuitive and personalized prediction tool.
Serum TSH is the most sensitive indicator re ecting thyroid function. With the continuous improvement of measurement technology, it has become the rst-line screening index for subclinical thyroid dysfunction [32]. TSH is secreted by the adenohypophysis and can promote the growth and differentiation of thyroid tissue and the synthesis and release of thyroid hormones. The TSH level is affected by various factors such as age, gender, genetics, geographic area, day-night cycle and measurement methods. Serum TSH level is not only an important factor in the development of thyroid nodules but TSH can also be used as an important indicator to judge the benign and malignant tumors of patients with thyroid nodules. Serum TSH is also the best indicator for screening and diagnosing hypothyroidism [31][32][33]. Most patients with thyroid nodules have normal thyroid function. Studies have shown [12] that elevated serum TSH levels are one of the risk factors for clinical hypothyroidism in patients with thyroid nodules within 20 years. In patients with thyroid nodules, as the serum TSH increases, the risk of thyroid malignant tumors also increases [34][35][36][37] .
In recent years, with the increasing number of patients with hypothyroidism in the population, the impact of iodine and uoride on thyroid function has become a hot issue of common concern but the research results are different [7,13,14] [15-21, 38, 39] . This study found that gender and urinary iodine are in uencing factors of TSH increase, which is consistent with the results of studies such as Vanderpump MP [12], Danchen Wang [40], and Wang D [41] Drinking water uoride is a critical in uencing factor. Some studies have found that [6,15,16,42,43] high uoride and high iodine and high uoride are related to elevated serum TSH. Malin AJ et al [14] found that there is a correlation between serum uoride and urine uoride and drinking water uoride-serum uoride, serum uoride and TSH. Therefore, we also regard drinking water as one of the in uencing factors.
Studies have shown that although some patients with elevated TSH can return to normal on their own [12,44] There are still some patients who will develop hypothyroidism [12], so early individualized prediction and treatment are very important. Of the 313 patients with thyroid nodules in this survey, 53 patients had serum TSH higher than the reference value and 4 patients had lower reference values.
Therefore, this study only explored the predictive model of TSH elevation in patients with thyroid nodules. According to the urinary iodine, gender, and drinking water uoride are the independent factors in uencing the increase of serum TSH concentration A Nomogram model with a predictive e cacy of 0.678 was established. The Nomogram model is intuitive and visual and can realize individualized prediction according to the value of each individual factor.
Due to the small sample content, this study only constructed the Nomogram and failed to carry out external veri cation. Due to the lack of data of water-source low iodine and high uorine population, this prediction model is not suitable for the prediction of serum TSH increase in water-source low iodine and high uorine population. In the future, further data should be collected to improve the prediction model.
In summary, the Nomogram prediction model established in this study allows doctors to assess the risk of elevated serum TSH in patients with asymptomatic thyroid nodules based on gender, urine iodine content and uoride content in drinking water. On the one hand, it can be determined whether the patient has a thyroid function test. On the other hand, high-risk groups with hypothyroidism can be detected early and corresponding measures can be taken to reduce the progression of thyroid nodules to hypothyroidism. It has high clinical application value and is worth promoting.

Conclusion
Gender, urine iodine and drinking water uoride of thyroid nodules are independent in uencing factors for the rise of serum TSH. The Nomogram can become a predictive accuracy and differentiation. The ow chart of the selected subjects.

Figure 2
Lasso coe cient pro les of the 6 features. A coe cient pro les plot was produced against the log (λ) sequence. Vertical line was drawn at the value selected using 10-fold cross-validation, where optimal l resulted in 4 nonzero coe cients. the AUC is to 1, the greater the predicted value of the model is.