A logistic regression model for predicting postoperative recurrence of cellular uterine leiomyoma

Yanling Zhong: Writing – review & editing, Writing – original draft, Visualization, Investigation, Formal analysis, Conceptualization. Mu Xu: Investigation, Data curation. Liangzhi Cai: Supervision, Project administration.

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Of 414 initially enrolled patients, 60 were excluded due to residual leiomyoma and 178 were lost to follow-up, leaving 176 for analysis. Baseline characteristics did not differ significantly between included and excluded patients ( Supplementary Table 1 ), suggesting that loss to follow-up was likely random. The final cohort was followed for a median of 4 years (range: 1.75–7.5 years). Recurrence occurred in 76 patients (43.18%). Recurrence time ranged from 6 to 84 months (mean: 30.8 months; the median recurrence time estimated by the Kaplan-Meier method was 24.0 months (95% CI: 24.0 – 36.0 months)）. Among recurrent cases, 10 underwent repeat surgery: four were confirmed as CUL, five as ordinary leiomyoma, and one as leiomyosarcoma. Except for one case with pelvic, abdominal, and sigmoid colon metastasis, all recurrences were uterine-localized. The malignancy rate of CUL was 0.57% (1/176). Univariate analysis identified no significant association between postoperative recurrence and age, parity, delivery history, prior leiomyoma surgery, preoperative symptoms, CA125, LDH, triglyceride, total cholesterol, high-density lipoprotein cholesterol, and low-density lipoprotein cholesterol levels, leiomyoma diameter, tumor type, ultrasound blood flow signals, or surgical approach ( p  > 0.05). Conversely, leiomyoma multiplicity, BMI, and endometriosis were significantly correlated with recurrence ( p  < 0.05; Table 1 ). Multivariate analysis identified ≥ 2 leiomyomas, coexisting endometriosis, and BMI ≥ 24 kg/m² as independent risk factors for CUL recurrence ( Table 2 ). Table 1 Univariate analysis of postoperative recurrence in myomectomy of Cellular Uterine Leiomyoma. Observation target Non-recurrence group (100 cases) Recurrence group (76 cases) X 2 OR (95% CI) P Age [cases (%)] 0.221 0.638 ≤ 40 57 (57.00) 46 (60.53) 1.00 (Reference) ＞40 43 (43.00) 30 (39.47) 0.86 (0.47 ∼ 1.59) Menopause [cases (%)] 0.486 0.486 No 94 (94.00) 74 (97.37) 1.00 (Reference) Yes 6 (6.00) 2 (2.63) 0.42 (0.08 ∼ 2.16) BMI [cases (%)] 13.591 < 0.001 * ＜24.0 Kg/m2 85 (85.00) 46 (60.53) 1.00 (Reference) ≥ 24.0 Kg/m2 15 (15.00) 30 (39.47) 3.70 (1.81 ∼ 7.56) Gestational number [cases (%)] 1.343 0.511 0 10 (10.00) 12 (15.79) 1.00 (Reference) 1 23 (23.00) 17 (22.37) 0.62 (0.22 ∼ 1.76) ≥ 2 67 (67.00) 47 (61.84) 0.58 (0.23 ∼ 1.46) Parity [cases (%)] 3.175 0.204 0 16 (16.00) 19 (25.00) 1.00 (Reference) 1 49 (49.00) 38 (50.00) 0.65 (0.30 ∼ 1.44) ≥ 2 35 (35.00) 19 (25.00) 0.46 (0.19 ∼ 1.09) History of previous fibroid resection [cases (%)] 0.138 0.71 No 97 (97.00) 72 (94.74) 1.00 (Reference) Yes 3 (3.00) 4 (5.26) 1.80 (0.39 ∼ 8.28) Endometriosis [cases (%)] 33.403 < 0.001 * No 91 (91.00) 40 (52.63) 1.00 (Reference) Yes 9 (9.00) 36 (47.37) 9.10 (4.01 ∼ 20.65) CA125 [cases (%)] 1.28 0.257 ＜35 U/ml 91 (91.00) 65(85.53) 1.00 (Reference) ≥ 35 U/ml 9 (9.00) 11 (14.47) 2.01 (0.73 ∼ 5.56) LDH [cases (%)] 0.057 0.822 ≤ 250 U/L 93 (93.00) 70 (92.11) 1.00 (Reference) ＞250 U/L 7 (7.00) 6 (7.89) 1.14 (0.37 ∼ 3.54) TG [cases (%)] 0.071 0.79 ＜2.26 mmol/L 94 (94.00) 73 (96.05) 1.00 (Reference) ≥ 2.26 mmol/L 6 (6.00) 3 (3.95) 0.64 (0.16 ∼ 2.66) TC [cases (%)] 0.881 0.348 ＜6.22 mmol/L 91 (91.00) 72 (94.74) 1.00 (Reference) ≥ 6.22 mmol/L 9 (9.00) 4 (5.26) 0.56 (0.17 ∼ 1.90) HDL-C [cases (%)] 0.127 0.721 ≥ 1.04 mmol/L 92 (92.00) 71 (93.42) 1.00 (Reference) ＜1.04 mmol/L 8 (8.00) 5 (6.58) 0.81 (0.25 ∼ 2.58) LDL-C [cases (%)] 1.323 0.25 ＜4.14 mmol/L 90 (90.00) 64 (84.21) 1.00 (Reference) ≥ 4.14 mmol/L 10 (10.00) 12 (15.79) 1.69 (0.69 ∼ 4.14) Number of uterine fibroids [cases (%)] 14.969 < 0.001 * 1 61 (61.00) 24 (31.58) 1.00 (Reference) ≥ 2 39 (39.00) 52 (68.42) 3.39 (1.81 ∼ 6.36) Maximum diameter of uterine fibroids [cases (%)] 0.858 0.651 ＜5 cm 10 (10.00) 10 (13.16) 1.00 (Reference) ≥ 5 cm-＜10 cm 82 (82.00) 58 (76.32) 0.71 (0.28 ∼ 1.81) ≥ 10 cm 8 (8.00) 8 (10.53) 1.00 (0.27 ∼ 3.72) Type of fibroids [cases (%)] 0.245# 1 Submucosal leiomyoma 3 (3.00) 3 (3.95) 1.00 (Reference) Intramural Leiomyoma 86 (86.00) 65 (85.53) 0.76 (0.15 ∼ 3.87) Subserosal leiomyoma 11 (11.00) 8 (10.53) 0.73 (0.12 ∼ 4.59) Surgical approach [cases (%)] 0.827 0.363 Laparoscopic surgery 73 (73.00) 60 (78.95) 1.00 (Reference) Open surgery 27 (27.00) 16 (21.05) 0.72 (0.36 ∼ 1.46) Abbreviations: BMI, body mass index; CA125, cancer antigen 125; LDH, lactate dehydrogenase; TG, triglyceride; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol. * p  < 0.05; # Kruskal–Wallis H test.Bold indicates statistical significance Table 2 Multivariate analysis of postoperative recurrence after myomectomy for cellular uterine leiomyomas. Variable β SE Z P OR (95% CI) BMI ＜24.0 Kg/m2 1.00 (Reference) ≥ 24.0 Kg/m2 1.43 0.42 3.43 < 0.001 4.17 (1.84 ∼ 9.42) Endometriosis No 1.00 (Reference) Yes 2.33 0.45 5.19 < 0.001 10.32 (4.27 ∼ 24.94) Number of uterine fibroids 1 1.00 (Reference) ≥ 2 1.02 0.37 2.77 0.006 2.77 (1.35 ∼ 5.71) Abbreviations: BMI, body mass index; CI, confidence interval; OR, odds ratio; SE, standard error. Univariate analysis of postoperative recurrence in myomectomy of Cellular Uterine Leiomyoma. Abbreviations: BMI, body mass index; CA125, cancer antigen 125; LDH, lactate dehydrogenase; TG, triglyceride; TC, total cholesterol; HDL-C, high-density lipoprotein cholesterol; LDL-C, low-density lipoprotein cholesterol. * p  < 0.05; # Kruskal–Wallis H test.Bold indicates statistical significance Multivariate analysis of postoperative recurrence after myomectomy for cellular uterine leiomyomas. Abbreviations: BMI, body mass index; CI, confidence interval; OR, odds ratio; SE, standard error. To predict postoperative recurrence of CUL, we developed a model incorporating three key factors: multiple leiomyomas (≥ 2), concurrent endometriosis, and elevated BMI (≥ 24 kg/m²) ( Fig. 1 ). The Hosmer-Lemeshow test confirmed model adequacy (χ² = 0.51, p  = .997), demonstrating strong alignment between predicted and observed outcomes. Fig. 2 , Fig. 3 , Fig. 4 present the ROC curve, calibration plot, and decision curve analysis for recurrence prediction. The model achieved an area under curve of 0.791 (95% CI: 0.724–0.858), with a sensitivity of 65.8% and specificity of 84.0%. Calibration analysis showed close agreement between the predicted and actual recurrence probabilities. Decision curve analysis indicated clinically meaningful predictive utility across the relevant threshold probabilities. Fig. 1 Nomogram for Predicting Postoperative Recurrence of Cellular Uterine Leiomyoma. The nomogram incorporates BMI and other predictive variables to estimate the probability of recurrence after surgery. BMI = body mass index. Fig. 2 Receiver operating characteristic curve. Fig. 3 Calibration curve. Fig. 4 Decision curve. Nomogram for Predicting Postoperative Recurrence of Cellular Uterine Leiomyoma. The nomogram incorporates BMI and other predictive variables to estimate the probability of recurrence after surgery. BMI = body mass index. Receiver operating characteristic curve. Calibration curve. Decision curve.

A total of 414 patients with CUL who underwent myomectomy at Fujian Maternal and Child Health Hospital between January 1, 2012, and December 31, 2021, were initially enrolled.Clinical and demographic data were collected from electronic medical records. Postoperative follow-up information was obtained through outpatient clinic visits and telephone interviews, with follow-up data updated until September 31,2023. Inclusion criteria: 1. Hospitalized patients who underwent myomectomy in the gynecology department. 2. Postoperative pathological confirmation of CUL. 3. Availability of complete clinical and pathological data. 4. Regular postoperative follow-up examinations. Hospitalized patients who underwent myomectomy in the gynecology department. Postoperative pathological confirmation of CUL. Availability of complete clinical and pathological data. Regular postoperative follow-up examinations. Exclusion criteria: (1) Residual lesions (n = 60): Patients with evidence of residual tumor on postoperative imaging (ultrasound or MRI performed within 3 months of surgery) were excluded, as our study aimed to predict recurrence after complete resection, and residual disease represents persistent disease rather than true recurrence (2) Concurrent malignant or borderline tumors. (3) Patients who simultaneously underwent hysteroscopic myomectomy.Patients who underwent hysteroscopic myomectomy were excluded because this technique differs substantially from transabdominal or laparoscopic approaches in lesion visualization and resection completeness, which may independently influence recurrence risk. (4) Lost to follow-up (n = 178): Patients who did not complete any follow-up visit beyond the initial 3-month postoperative assessment, despite documented attempts at contact via phone and mail, were considered lost to follow-up and excluded. Residual lesions (n = 60): Patients with evidence of residual tumor on postoperative imaging (ultrasound or MRI performed within 3 months of surgery) were excluded, as our study aimed to predict recurrence after complete resection, and residual disease represents persistent disease rather than true recurrence Concurrent malignant or borderline tumors. Patients who simultaneously underwent hysteroscopic myomectomy.Patients who underwent hysteroscopic myomectomy were excluded because this technique differs substantially from transabdominal or laparoscopic approaches in lesion visualization and resection completeness, which may independently influence recurrence risk. Lost to follow-up (n = 178): Patients who did not complete any follow-up visit beyond the initial 3-month postoperative assessment, despite documented attempts at contact via phone and mail, were considered lost to follow-up and excluded. After applying these criteria, 176 patients were included in the study. To evaluate whether loss to follow-up introduced selection bias, we compared baseline demographic and clinical c haracteristics between patients included in the final analysis (n = 176) and those lost to follow- up (n = 178). The results are presented in Supplementary Table 1 .Based on postoperative recur rence status, patients were categorized into a recurrence group (n = 76) and a non-recurrence group (n = 100). This study strictly adheres to the principles of the Declaration of Helsinki. Since this study used fully anonymized data without any risk to participant privacy, the Ethics Committee of Fujian Maternity and Child Health Hospital granted an exemption from informed consent, and the study protocol was approved (approval No. 2024KY042). Clinical and pathological data included age, menopausal status, body mass index (BMI), pregnancy history, parity, prior myomectomy, cancer antigen 125 (CA125) level, lactate dehydrogenase (LDH) level, lipid profile, endometriosis comorbidity, tumor number, maximum tumor diameter, and tumor type.All laboratory variables, including CA125, were measured from blood samples obtained within 7 days prior to surgery. Other clinical and demographic data were recorded at hospital admission. (1)Recurrence was defined as the detection of new leiomyomas ≥ 1 cm in diameter on transvaginal ultrasound or magnetic resonance imaging performed at least 6 months after surgery, in patients who had complete resection confirmed by postoperative imaging [11] (2)For clinical interpretability, age was dichotomized at 40 years, as this cutoff is commonly used in gynecological research to approximate the transition in reproductive aging [12] . This categorization is consistent with previous studies on uterine fibroid recurrence [13] and maintained adequate subgroup sample sizes for analysis. Data processing and analysis were performed using SPSS 26.0, R version 4.3.3, and Zstats 1.0 ( www.zstats.net ). Categorical variables were expressed as frequencies (percentages) and compared using χ² tests. Univariate logistic regression was used to evaluate the association between each potential risk factor and postoperative recurrence, with odds ratios (ORs) and 95% confidence intervals (CIs) calculated. Variables with p < 0.05 in univariate analysis were entered into multivariate logistic regression to identify independent risk factors (two-tailed, α = 0.05). The significant predictors from multivariate analysis were used to develop a predictive model using the “Zstats v1.0” package in R. Model performance was assessed using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis.

Postoperative recurrence of CUL is a common challenge encountered by both patients and doctors. At present, the high-risk factors and pathogenic mechanisms underlying CUL recurrence remain unclear. In this institutional study, we found that BMI ≥ 24 kg/m², concomitant endometriosis, and multiple leiomyomas (≥ 2) are independent risk factors for postoperative recurrence. We constructed a user-friendly nomogram prediction model based on these clinical characteristics to help clinicians assess recurrence risk preoperatively. The model allows for targeted therapy and a personalized follow-up protocol, aiming to reduce recurrence and improve postoperative quality of life in women.

Uterine fibroids affect many women, with recurrence rates of 10–40% and malignant transformation in < 1% [14] , [15] . Patients suffer from depression, bleeding, and pain, impairing quality of life [16] , [17] . CUL, the most common special subtype, shares hormonal and hypoxic regulatory mechanisms with ordinary fibroids [18] , [19] , but also exhibits genetic features (e.g., 1p deletion, HMGA2 rearrangements, MED12 mutations) similar to leiomyosarcoma, suggesting malignant potential [5] , [20] . Taran et al. proposed that CUL, benign metastasizing leiomyoma, and leiomyosarcoma may share pathogenic mechanisms [6] . However, few studies have focused on CUL recurrence. This study provides a clinical reference for its prevention. This study identified multiple leiomyomas (≥ 2), coexisting endometriosis, and a BMI of ≥ 24 kg/m 2 as independent predictors of recurrence following conservative resection of CUL.Multiple fibroids increased recurrence risk, consistent with previous reports [15] . A higher fibroid number raises the likelihood of missed small lesions during uterus-preserving surgery, which may subsequently grow. Khamaiseh et al. similarly found that the number of removed fibroids predicted reintervention (HR 1.21) and highlighted genetic predisposition (e.g., FH, YEATS4) as contributing factors [21] .BMI ≥ 24 kg/m² was independently associated with recurrence. Obesity may promote recurrence by creating a pro-tumor microenvironment via chronic low-grade inflammation, elevated inflammatory markers (e.g., SII), and increased adipocyte aromatase activity driving estrogen production [22] , [23] . This aligns with multifactorial models; for instance, C. Song et al. [24] incorporated clinical predictors into a nomogram for post-myomectomy recurrence, highlighting that factors such as fibroid subtype and residual disease interact with underlying biological processes.Future models integrating inflammatory biomarkers (e.g., IL−6) could refine risk stratification for CUL.The higher recurrence risk in CUL patients with coexisting endometriosis may stem from shared hormonal dependence. Ectopic endometrium produces estrogen and inflammatory factors (e.g., IL−6, TNF-α), creating a proliferative microenvironment [25] , [26] . Additionally, surgical treatment of endometriosis might disseminate malignant cells, especially during laparoscopy.These findings underscore the multifactorial nature of CUL recurrence, involving clinical factors (fibroid number, BMI, endometriosis), molecular susceptibility, and inflammatory pathways. Surgical approach (open vs. laparoscopic myomectomy) did not significantly influence recurrence in our study, though this finding requires careful interpretation. Historically, laparoscopy was thought to carry higher recurrence risk due to lack of tactile feedback [27] . However, recent evidence challenges this: Otten et al. reported lower recurrence after laparoscopy (10.2% vs. 23.8%) [28] , and Äyräväinen et al. found comparable quality of life regardless of approach [29] . Advances in laparoscopic technology (better visualization, energy devices, and techniques) have narrowed the gap in resection completeness [30] . Surgeon expertise also plays a critical role; better outcomes are often reported from high-volume centers [31] , [32] , while historical studies may reflect learning curves or older technology. Our single-institution design may not fully capture variations in surgeon experience. Furthermore, confounding by indication (larger/more numerous fibroids selected for laparotomy) cannot be excluded despite adjustment. Although we did not directly document morcellator use, the high CUL recurrence rate (43.18%)—with 10 reoperations and one malignant transformation—suggests that residual tumor fragments are a key mechanism. Clinicians should carefully review preoperative imaging and avoid morcellators in suspicious cases or use closed techniques.Overall, while surgical approach may not be a major determinant of CUL recurrence, patient selection, technology, and surgeon expertise remain paramount. Future multicenter studies are needed to validate these findings. To inform clinical decision-making, we constructed a nomogram translating the three predictors (BMI≥24 kg/m², endometriosis, multiple fibroids) into individualized recurrence risk estimates.The following scenarios demonstrate how this tool might guide postoperative management in practice.Scenario 1 (High-risk, score 200, risk >90%): For patients with BMI≥ 24 kg/m², endometriosis, and multiple fibroids, consider frequent surveillance (e.g., ultrasound every 6 months for 2 years), adjuvant therapy (GnRH agonizts, SPRMs), or hysterectomy if childbearing complete.Scenario 2 (Low-risk, score 0, risk <30%): For patients with BMI< 24 kg/m², no endometriosis, and solitary fibroid, standard follow-up (annual ultrasound) is sufficient; uterus-preserving surgery can be offered confidently.Scenario 3 (Intermediate-risk, score 100, risk 50–60%): For patients with BMI ≥ 24 and multiple fibroids but no endometriosis, enhanced surveillance (e.g., ultrasound every 6–12 months) is recommended without immediate adjuvant therapy.Shared decision-making should balance fertility desires and definitive options.These examples demonstrate how the nomogram stratifies recurrence risk and guides individualized surveillance and adjuvant therapy. Furthermore, it should be noted that patients undergoing hysteroscopic myomectomy were excluded from this study. Therefore, our predictive model is primarily applicable to patients undergoing laparoscopic or open myomectomy. Caution is warranted when generalizing this model to all patients undergoing uterus-preserving surgery.Additionally, the median follow-up duration of our cohort was 48 months, which may not be sufficient to capture late recurrences occurring beyond 5 years. Future prospective studies with longer follow-up periods are needed to better characterize the long-term recurrence patterns of cellular uterine leiomyoma In summary, clinical decisions should integrate a thorough assessment of the patient's overall health, systemic factors, leiomyoma features, reproductive goals, and operative risks. For patients suspected of having CUL—especially those with BMI ≥ 24 kg/m², concomitant endometriosis, or multiple leiomyomas—uterus-preserving surgery should be selected with caution. Postoperative follow-up should be strengthened, with particular attention paid to the early signs of recurrence in patients at high risk. For patients at a high risk of recurrence, adjuvant drug therapy (e.g., GnRH-a or selective progesterone receptor modulators) may be considered to reduce the risk of recurrence.

Several limitations of this study should be acknowledged.First, the retrospective, single-center design may introduce selection bias. The exclusion of 57% of the initial cohort, primarily due to loss to follow-up, is a significant limitation. However, baseline characteristics did not differ significantly between included and excluded patients ( Supplementary Table 1 ), suggesting attrition was likely random. Nevertheless, unmeasured factors (e.g., socioeconomic status, healthcare access) may have influenced compliance and recurrence risk. Future prospective multicenter studies with standardized protocols, active patient engagement (e.g., reminder systems), and linkage to national registries are needed. External validation in independent cohorts with longer follow-up is also warranted.

Uterine fibroids affect > 70% of women globally [1] . Cellular uterine leiomyoma (CUL) is the most common histological subtype, accounting for 17.9% of all fibroids, with an average recurrence time of ∼28.6 months [2] , [3] , [4] . Some CULs exhibit borderline features, including distant metastasis and malignant transformation [5] , [6] , [7] . Myomectomy is preferred for young or fertility-desiring patients, but uterus-preserving surgery carries higher recurrence risk due to CUL's biological behavior. CUL recurrence has been reported even after total hysterectomy [9] , and metastatic cases exist [8] . Few studies have specifically investigated CUL recurrence risk factors. Rothmund et al. [3] reported a 12% recurrence rate after uterus-conserving surgery.Ming et al. [10] found that CUL patients more often have concurrent adenomyosis and multiple fibroids. However, no predictive model exists for CUL, and risk factors remain unclear. This study aimed to identify risk factors for postoperative CUL recurrence and develop a clinical predictive model to support treatment decisions.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.