Abstract
Due to the unclear mechanism of uterine fibroids, more risk factors need to be clarified. This study explored associated factors of uterine fibroids and established a prediction model using clinical data from our institution. Logistic and Least Absolute Shrinkage and Selection Operator (LASSO) analyses were used to screen the key associated factors of uterine fibroids. The receiver operating characteristic (ROC) curve and DeLong test were used to analyze the prediction performance of indicators. XGBoost classification and random forest were used to rank the feature importance. A prediction model based on the key factors was established. Decision Curve Analysis (DCA), ROC, and nomogram analysis were used to assess the performance of the model for predicting uterine fibroids. Of the 303 patients enrolled, 201 had uterine fibroids. Logistic and LASSO regression analyses identified five core risk indicators, including age, thyroid-stimulating hormone index (TSHI), number of deliveries, abnormal menstruation, and polycystic syndrome. Both ROC and feature importance ranking analyses consistently implied the importance of age and TSHI on uterine fibroid risk. With the increase of age and TSHI, the risk of uterine fibroids was significantly increased (all p for trend <0.05). The combination of age and TSHI achieved favorable performance and clinical net benefit in fibroids risk prediction, and its favorable performance was also validated in the external National Health and Nutrition Examination Surveys database. Age and TSHI were the key associated factors of uterine fibroids, and their combination had promising clinical value for predicting the risk of uterine fibroids.
Introduction
Uterine fibroids are the most common benign tumors of the female reproductive system, consisting of smooth muscle and connective tissue. Uterine fibroids occur mainly in the 35–55 age group. Most of the patients have no obvious symptoms in the pre-tumor stage, and in the later stage, gynecological symptoms such as menstrual disorders, lower abdominal masses, and pelvic pressure symptoms. 1 It has been reported that about 25% of women with uterine fibroids have severe symptoms by the time they receive treatment, 2 and uterine fibroids are one of the main reasons for hysterectomies in women, so they may have a more serious impact on women’s reproductive health, mental health, and quality of life.
In 2019, the age-standardized incidence of uterine fibroids worldwide was 241.18 per 100,000. Among the countries with the highest incidence rates were Latvia, the Russian Federation, and Ukraine, with rates of 667.14 per 100,000, 586.64 per 100,000, and 578.21 per 100,000, respectively. While New Zealand, Australia, and North Korea exhibited lower incidence rates, with 81.98 per 100,000, 86.13 per 100,000, and 103.72 per 100,000, respectively. 3
Up to now, the pathogenesis of uterine fibroids is not completely clear. A number of studies have indicated that genetic factors, sex hormones, and their receptors play a significant role in the formation and growth of uterine fibroids.4,5 In a two-sample Mendelian random analysis conducted by Wang et al., the relationship between female reproductive factors, sex hormones and the risk of uterine fibroids was investigated. 6 The results indicated that menopausal age, the number of live births, and the total testosterone level had a causal relationship with the risk of uterine fibroids. Due to the unclear mechanism, more risk factors need to be clarified.
This study explored the risk factors of uterine fibroids using the data of the inpatients admitted to our hospital, which was beneficial to the understanding of patients with uterine fibroids, and provided a practical basis for healthcare professionals and patients to effectively reduce the incidence of uterine fibroids, ultimately achieving the goal of improving physical and mental health and enhancing the quality of life.
Significance of this Study
Uterine fibroids may have a more serious impact on women’s reproductive health, mental health, and quality of life.
This study explored the risk factors of uterine fibroids using the data of the inpatients admitted to our hospital, which was beneficial to the understanding of patients with uterine fibroids.
Our study emphasized the importance of age and TSHI in uterine fibroids and provided reference value for similar research. At the same time, it is conducive to the medical staff to judge the state of the patient and make a better-quality decision.
Methods
Study participants
In this study, the patients hospitalized in the gynecology department of our hospital were included as study participants from January 2022 to August 2024. The fibroid group included female inpatients aged ≥18 years with uterine fibroids diagnosed by imaging modalities (ultrasonography, CT, or MRI) or confirmed by intraoperative and postoperative histopathological examination, regardless of fibroid size or surgical status. The control group included age-matched female inpatients hospitalized for other benign gynecological conditions (such as benign ovarian cysts, tubal factors, benign cervical lesions, ectopic pregnancy, or pelvic inflammatory disease) who were confirmed to have no uterine fibroids by imaging or surgical exploration. The exclusion criteria for both groups were as follows: (1) malignant or borderline gynecological tumors, including uterine sarcoma, endometrial cancer, cervical cancer, or ovarian malignancies; (2) severe cardiac, hepatic, renal, or coagulation disorders; (3) active infections, autoimmune diseases, connective tissue diseases, or severe endocrine disorders; (4) prior uterine artery embolization, high-intensity focused ultrasound, or radiofrequency ablation for fibroids; (5) history of uterine adenomyosis (control group only); (6) cognitive or psychiatric disorders precluding study participation; (7) incomplete clinical data; or (8) concurrent participation in other clinical trials that may interfere with the study variables.
A total of 303 patients were ultimately enrolled in the study, comprising 201 patients in the uterine fibroid group (diagnosed via postoperative histopathological examination (n = 105) or imaging modalities (n = 96)) and 102 patients in the control group.
Indicators
The indicators included in this study were age, BMI, albumin, propionate aminotransferase, bilirubin, uric acid, platelet, neutrophil count, lymphocyte count, neutral lymphatic ratio, platelet lymphatic ratio, hemoglobin, TSH, FT3, FT4, TT4R1, thyroid-stimulating hormone index (TSHI), prolactin, estrogen, pregestational hormone, CA125, alpha-fetoprotein, carcinoembryonic antigen, CA199, CA153, age of first birth, number of deliveries, number of miscarriages, marital status (married, other), education (middle school and below, high school, college degree or above), pregnancy, normal menstruation, birth history, polycystic syndrome, HPV infection, history of abortion, anemia.
The formula for calculating the TT4RI and TSHI was as follows:
Statistical analysis
The R language (version 4.3.0, http://www.R-project.org) was used for data analysis. The measurement data of non-normal distribution were represented by median (P25–P75), and tested by the Wilcoxon rank sum test; the measurement data of normal distribution were represented by mean ± standard deviation and tested by analysis of variance. The counting data were expressed by quantity (rate) and tested by the chi-square test.
A two-step variable selection strategy was adopted in this study. First, univariate logistic regression was used to screen out variables with statistically significant differences (p < 0.05) between the two groups, thereby excluding clearly irrelevant variables and reducing noise from variables with no apparent association with the outcome. Subsequently, Least Absolute Shrinkage and Selection Operator (LASSO) regression analysis was applied to the pre-screened variables to further address potential multicollinearity among the remaining candidates and to identify the most parsimonious and robust subset of predictors. This sequential approach combining univariate pre-screening, LASSO regularization, and multivariate logistic regression has been adopted in several clinical prediction model studies.7,8 Variance inflation factor (VIF) analysis was performed to assess multicollinearity among the variables included in the multivariate logistic regression model. A VIF value <5 was considered indicative of no significant multicollinearity.
The receiver operating characteristic (ROC) curve was used to analyze the prediction performance of different indicators, and the DeLong test was used to compare the prediction performance of different indicators. XGBoost classification and random forest were used to rank the feature importance of indicators. Trend regression was used to explore the correlation between age, TSHI and the risk of uterine fibroids. Decision Curve Analysis (DCA) was used to evaluate the application value of the prediction model in actual clinical decision-making. A nomogram was used to analyze the risk of uterine fibroids by combining the two indices. Finally, we conducted an external validation on the model for predicting the uterine fibroids based on National Health and Nutrition Examination Surveys (NHANES) data (2001–2002 cycle). p < 0.05 was considered to be statistically significant.
Results
Patient baseline data
A total of 303 patients were enrolled in this study, of which 201 patients had uterine fibroids and 102 did not. The median ages in the control and uterine fibroid groups were 34.000 [26.000, 42.000] and 47.000 [41.000, 51.000], respectively. As can be seen from Table 1, compared to the control group, patients in the fibroid group were older, had a higher lymphocyte count, a higher platelet lymphatic ratio, higher levels of FT4, TT4R1, TSHI, and lower levels of prolactin, estrogen, pregestational hormone (all p < 0.05). In addition, patients in the uterine fibroid group had higher levels of alpha-fetoprotein, a higher number of deliveries and miscarriages. There were differences in marital status, education, pregnancy, normal menstruation, birth history, polycystic syndrome, history of abortion, and anemia between the two groups.
Patient baseline data.
TSHI: thyroid-stimulating hormone index.
The associated factors of uterine fibroids
The variables showing the differences (p < 0.05) between the control group and the uterine fibroids groups were selected for conducting association analysis. Results (Table 2) of univariate logistic regression suggested that 13 variables, including age, platelet lymphatic ratio, TSHI, prolactin, number of deliveries, number of miscarriages, marital status, education, abnormal menstruation, birth history, polycystic syndrome, history of abortion, and anemia, were associated factors of uterine fibroids. To further address potential multicollinearity among these variables and select the most robust predictors, LASSO regression was subsequently performed. At the optimal penalty parameter (λ = 0.009, minimum mean square error), the LASSO model retained 10 associated factors: age, TSHI, prolactin, number of deliveries, number of miscarriages, marital status, abnormal menstruation, birth history, polycystic syndrome, and anemia (Figure 1(a), (b)). These 10 variables were then entered into a multivariate logistic regression model to identify independently associated factors. The VIF values for all 10 variables are all below the threshold of 5, confirming that multicollinearity was not a significant concern in the final model (Supplemental Table 1). As shown in Table 3, age, TSHI, number of deliveries, abnormal menstruation, and polycystic syndrome were independently related to the uterine fibroids (all p < 0.05).
Screening influencing factors of uterine fibroids by univariate logistic regression.
TSHI: thyroid-stimulating hormone index.
LASSO regression was conducted to remove the invalid factors associated with uterine fibroids. (a) Variation trend of LASSO regression coefficient with λ parameter (1: marital status, 2: education, 3: number of miscarriages, 4: birth history, 5: polycystic syndrome, 6: education, 7: anemia, 8: age, 9: platelet lymphatic ratio, 10: TSHI, 11: prolactin, 12: number of miscarriages, 13: number of deliveries), (b) Variation trend of binomial deviance with λ parameter in LASSO regression model.
Screening influencing factors of uterine fibroids by multivariate logistic regression.
TSHI: thyroid-stimulating hormone index.
Predictive performance of a single influencing factor
We then used ROC to analyze the prediction performance of five independent indicators on uterine fibroids. As can be seen from Table 4, the area under curve (AUC) of age, TSHI, number of deliveries, normal menstruation, and the polycystic syndrome were 0.801, 0.668, 0.602, 0.621, and 0.541 respectively. ROC analysis showed that age had the most favorable prediction performance, followed by TSHI, which highlighted their importance in uterine fibroids. The results of the DeLong test (Table 5) indicated that the prediction performance of age was better than other indicators, while the predictive performance of polycystic syndrome was worse than that of other indicators.
Predictive performance of a single influencing factor on uterine fibroids risk.
AUC: area under curve; TSHI: thyroid-stimulating hormone index.
Comparison of the predictive performance of a single influencing factor on uterine fibroids.
TSHI: thyroid-stimulating hormone index.
In addition, we also assessed their value by performing feature importance analysis using two machine learning algorithms, and two algorithms commonly showed (Figure 2) that age and TSHI ranked first and second place, respectively. ROC and feature importance analysis all highlighted the importance of age and TSHI.
Ranking of feature importance of associated factors. (a) XGBoost classification, (b) Random forest.
We further used trend regression analysis to explore the association of age and TSHI changes with the risk of uterine fibroids. From Table 6, with the increase of age, the risk of uterine fibroids increases (p for trend <0.001). The same trend was found in TSHI (p for trend <0.001), especially when TSHI was greater than 0.824, the risk of uterine fibroids was dramatically increased (OR = 23.912, p < 0.001).
The association of age and TSHI with the risk of uterine fibroids.
Adjusting number of deliveries, abnormal menstruation, and polycystic syndrome.
TSHI: thyroid-stimulating hormone index.
Predictive performance of the comprehensive model on uterine fibroids
Based on the above results, age and TSHI were regarded as the key factors associated with the risk of uterine fibroids. Further, the prediction performance of age combined with TSHI was explored in this study. As observed from Table 7, the AUC of model 1 (age + TSHI) in the training and validation sets was 0.842 and 0.818, respectively. After considering the other three variables (model 2), the AUC was increased to 0.855 in the training set and 0.824 in the validation set, respectively. ROC analysis indicated that the combination of age and TSHI has achieved a relatively favorable performance.
Predictive performance of the comprehensive model on uterine fibroids risk.
AUC: area under curve; TSHI: thyroid-stimulating hormone index.
Model 1: age + TSHI, model 2: age + TSHI + number of deliveries + abnormal menstruation + polycystic syndrome.
The results of DCA (Figure 3(a)) demonstrated that across a threshold probability range of approximately 0.06–0.95, using the prediction model to guide enhanced surveillance decisions yielded a higher net benefit compared with the strategies of screening all patients or screening none, suggesting that the model has potential clinical utility in identifying women who may benefit from more frequent gynecological monitoring. Similar results were also found in model 2 (Figure 3(b)). DCA analysis suggested a similar clinical net benefit between model 1 and model 2.
Net benefit of different models for predicting the uterine fibroids. (a) DCA of model 1, (b) DCA of model 2. Model 1: age + TSHI, model 2: age + TSHI + number of deliveries + abnormal menstruation + polycystic syndrome. The x-axis represents the threshold probability, defined as the minimum predicted risk at which enhanced surveillance would be recommended. The y-axis represents the net benefit. The clinical decision evaluated here is whether to recommend enhanced gynecological surveillance, including more frequent ultrasonographic monitoring and earlier referral to a gynecologist, or women identified as high-risk by the model. The black line represents the “treat none” strategy (no enhanced surveillance for any patient), and the gray line represents the “treat all” strategy (enhanced surveillance for all patients). The red line represents the prediction model. Across a clinically relevant threshold probability range of approximately 0.06–0.95, the prediction model demonstrated a higher net benefit compared with both the “treat all” and “treat none” strategies, indicating that using the model to guide surveillance decisions would result in more clinical benefit than either default strategy.
Both ROC and DCA indicated the superiority of the combination of age and TSHI for predicting uterine fibroids. Then, age and TSHI were used to construct a nomogram model for predicting the risk of uterine fibroids (Figure 4(a)). The calibration curve (Figure 4(b)) showed that the bias-corrected line in the calibration graph was close to the ideal line, suggesting that the predicted results of the model were in good agreement with the actual results.
The risk assessment of uterine fibroids by combining age and TSHI. (a) Nomogram, (b) calibration curve.
External validation of the optimal model for predicting uterine fibroids based on NHANES data
Finally, we conducted an external validation on the optimal model for predicting the uterine fibroids based on NHANES data. The 1999–2006 NHANES cycles reported the data on uterine fibroids, but only 2001–2002 cycle recorded the FT4 and TSH data (for calculating TSHI). Therefore, we just enrolled the participants in the NHANES 2001–2002 cycle for the external validation analysis. This cycle had 11039 participants, a total of 229 participants met our inclusion criteria (female, age 18 or more, not pregnant, had FT4 and TSH data, had a uterine fibroids diagnosis record), of whom 34 participants were diagnosed with uterine fibroids.
The above analysis had indicated the superiority of the combination of age and TSHI for predicting uterine fibroids. External NHANES data further confirmed the favorable prediction performance of their combination (Figure 5(a), AUC = 0.83). We also performed the five-fold cross-validation to assess its stability, and the results showed the stable prediction performance both in training set and validation set (Table 8). These results fully demonstrated that their combination had good and stable predictive performance on uterine fibroids. At the same time, a favorable net benefit can also be obtained (Figure 5(b)).
External validation on optimal model (age + TSHI) for predicting the uterine fibroids based on NHANES data. (a) ROC analysis. (b) Net benefit analysis by DCA.
External validation of the optimal model for predicting uterine fibroids based on NHANES data.
AUC: area under curve; NHANES: national health and nutrition examination surveys.
Discussion
The incidence of uterine fibroids is high, affecting more than 70% of women in the world, especially Black women. 9 The etiology of uterine fibroids is not yet completely clear, but with further exploration in the fields of endocrinology and molecular science of uterine fibroids, people have gained a greater and deeper understanding and knowledge of their risk factors, and have found that their occurrence and development are affected by a combination of multiple factors, which involves many factors, such as biological characteristics, lifestyle and external environment, and so on. In this study, we found age, TSHI, number of deliveries, abnormal menstruation, and polycystic syndrome were associated factors of uterine fibroids.
Uterine fibroids do not occur before puberty, and their incidence increases with the increase in childbearing age. A prospective study found that the incidence of uterine fibroids increased with age (p ≤ 0.0001). The incidence was 6 percent in the 23–25 age group, 11 percent in the 29–31 age group, and 13 percent in the 32–35 age group. 10 We speculated that age was the most important risk factor for uterine fibroids, which may be related to the increasing age of the patient, accumulation of endogenous estrogens, and changes in the immune system. It is worth noting that age, as a surrogate for cumulative reproductive hormone exposure, may have indirectly captured the effects of estrogen and progesterone on fibroid risk, which could partially explain why these individual hormones were not retained in the final model despite their well-established biological roles in fibroid pathogenesis. In this study, blood samples were collected upon hospital admission without standardization for menstrual cycle phase. Given that serum levels of estrogen and progesterone fluctuate considerably across the menstrual cycle, 11 a single cross-sectional measurement may not adequately reflect long-term hormonal exposure, likely attenuating the observed associations. Moreover, the relatively modest sample size (n = 303) may have further limited the statistical power to detect moderate associations for hormones with high physiological variability.
For the first time, we found that TSHI is directly proportional to the risk of uterine fibroids. The higher the value of TSHI, the lower the sensitivity of thyroid hormone. TSHI is a thyroid hormone sensitivity index, derived from the mathematical modeling of the TSH–FT4 feedback relationship. Due to the complex network regulation of the hypothalamic–pituitary–thyroid (HPT) axis, composite indices such as TSHI can provide a more comprehensive and objective indication of thyroid hormone homeostasis than a single hormone level. 12 Therefore, some subtle alterations in thyroid axis regulation that remain undetectable by conventional single-parameter measurements may be captured by this integrated index. 13 This may explain why TSHI, rather than individual thyroid measures (TSH, FT3, FT4, or TT4R1), was identified as the only significant thyroid-related factor in our analysis.
Thyroid abnormalities are common in patients with uterine fibroids. 14 Yuk et al. discovered that uterine fibroid patients had an elevated risk of developing goiter and thyroid nodules. 15 Regarding their potential connected network, thyroid dysfunction and uterine fibroids are closely related to estrogen and have internal relations with the hypothalamus-pituitary-gland axis, which have similar physiological and pathological bases. Studies have demonstrated that estrogen may serve as a common link between the two conditions, as it regulates both uterine fibroid growth and thyroid function. Estradiol and progesterone promote the growth and enlargement of uterine fibroids. 16 Uterine fibroids are an estrogen-dependent condition, as they express higher levels of estrogen receptors (ER) and progesterone receptors compared to normal myometrium, demonstrating estrogen sensitivity.15,17 Estrogen promotes the growth of thyroid cells via two receptor subtypes, ERα and ERβ, thereby contributing to the development of thyroid nodules. 18 Upon binding to either ER-α or ER-β, estrogen forms an estrogen-ER complex, which translocates into the nucleus. Subsequently, receptor homodimerization or heterodimerization occurs, followed by binding to estrogen response elements (EREs) in the cell nucleus, ultimately influencing cellular growth and development. 19
In this study, it was found that the number of deliveries was a protective factor for uterine fibroids. Many studies have found that the number of deliveries is inversely proportional to the risk of uterine fibroids.20,21 Compared with parturient women, parturient women have a lower risk of uterine fibroids. Relevant studies have pointed out that the diameter of fibroids remains the same or becomes smaller after 6 months after delivery. 22 A case report indicated that childbirth has the potential to obstruct the blood supply to uterine fibroids, thereby reducing the size of the tumor. 23 At present, little is known about the protective mechanism of fertility for uterine fibroids. Some scholars believe it is related to hypoxia, apoptosis and other mechanisms. 24
In this study, abnormal menstruation was a risk factor for uterine fibroids. The relationship between uterine fibroids and menstrual cycle patterns is not clear. Some studies have found that the menstrual cycle has nothing to do with uterine fibroids. 25 A cross-sectional study conducted with a cohort of Korean nurses revealed an inverse relationship between irregular menstrual cycles during early adulthood and the subsequent occurrence of uterine fibroids. 26 We suspected that this phenomenon may be related to hormones.
Polycystic syndrome is a common disease caused by complex endocrine and metabolic abnormalities in women of childbearing age is characterized by chronic anovulation (disorder or loss of ovulation function) and hyperandrogenism (excess production of male hormones in women). Related studies have pointed out that polycystic ovary syndrome can reduce the risk of uterine fibroids. 27 Our research also found similar results. The relationship between them may be related to progesterone.
In addition to the above factors, other factors such as obesity and vitamin D level are also associated with uterine fibroids. A meta-study pointed out that the relationship between obesity and uterine fibroids was nonlinear and showed an inverted J-shaped pattern. 28 The level of vitamin D in patients with uterine fibroids decreased significantly. 29 A case-control study found that serum vitamin D level was negatively correlated with the volume of uterine fibroids. 30
It is important to acknowledge the limitations of this study. First, the relatively modest sample size (n = 303) from a single center may limit the generalizability of the findings. Although external validation using NHANES data (n = 229) demonstrated acceptable discriminative performance, future multi-center prospective studies with larger cohorts are warranted to confirm these results. Second, the control group consisted of patients hospitalized for other benign gynecological conditions rather than healthy community-based women. These patients may share certain risk factors with the fibroid group, such as hormonal profiles or reproductive histories, potentially leading to an underestimation of the true associations. Third, the fibroid group included both surgically confirmed cases (n = 105) and imaging modalities-diagnosed cases (n = 96), which may represent different disease severities. Small or deeply located fibroids may have been missed, potentially leading to misclassification bias; however, such misclassification would likely bias the results toward the null, making the observed associations more conservative. Fourth, blood samples were collected upon hospital admission without standardization for menstrual cycle phase, which may have introduced variability in serum reproductive hormone levels and attenuated their associations with uterine fibroids. Fifth, data on genetic factors associated with uterine fibroids were not collected, which may have influenced the results. Finally, XGBoost and Random Forest were employed solely as supplementary tools for variable importance ranking rather than for prediction, and the importance rankings should be interpreted as exploratory, given the limited sample size. Future studies with larger samples, standardized hormone sampling protocols, community-based controls, and longitudinal designs are needed to validate and extend these findings.
Conclusions
Age, TSHI, number of deliveries, abnormal menstruation, and polycystic syndrome were associated factors of uterine fibroids. The combination of age and TSHI may serve as a practical and preliminary risk stratification tool for uterine fibroids, and further validation in larger, prospective cohorts is needed to confirm its clinical utility.
Supplemental Material
sj-docx-1-imj-10.1177_10815589261451184 – Supplemental material for Analysis of risk factors for uterine fibroids and construction of prediction model
Supplemental material, sj-docx-1-imj-10.1177_10815589261451184 for Analysis of risk factors for uterine fibroids and construction of prediction model by Yapei Lu and Ziqi Cheng in Journal of Investigative Medicine
Footnotes
Ethical considerations
The study was approved by the Ethics Committee of First People’s Hospital of Fuyang District. All patients who were familiar with the contents and processes of the study and able to complete all the scheduled study processes signed the informed consent. Our study complies with the Declaration of Helsinki.
Author contributions
Yapei Lu contributed to the conception and design. Yapei Lu and Ziqi Cheng contributed to the collection and assembly of data. Yapei Lu and Ziqi Cheng analyzed and interpreted the data. All authors wrote and approved the final manuscript.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.
Declaration of conflicting interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Data availability statement
The data to support the findings of this study are available in the corresponding publications as given in the article.
Supplemental material
Supplemental material for this article is available online.
References
Supplementary Material
Please find the following supplemental material available below.
For Open Access articles published under a Creative Commons License, all supplemental material carries the same license as the article it is associated with.
For non-Open Access articles published, all supplemental material carries a non-exclusive license, and permission requests for re-use of supplemental material or any part of supplemental material shall be sent directly to the copyright owner as specified in the copyright notice associated with the article.
