Development and validation of a nomogram-based predictive model for recurrence risk of uterine leiomyoma after myomectomy.

This study was designed as a retrospective cohort analysis to identify factors associated with uterine fibroid recurrence after myomectomy and to develop a predictive model. Clinical data were collected from patients treated at the Affiliated Hospital of Guangdong Medical University. A nomogram was constructed based on statistically significant predictors to estimate individual recurrence risk. A total of 678 patients who underwent myomectomy between October 2015 and October 2022 were retrospectively included. Informed consent was obtained from all participants, and personal identifiers were removed in accordance with the Declaration of Helsinki. Data were extracted from the hospital’s electronic medical record system. Collected variables included demographic characteristics (age at surgery, BMI, age at menarche, parturition history, family history), pelvic comorbidities (adenomyosis, adenomyoma, endometriosis, benign adnexal masses), and fibroid-related ultrasound parameters (number of fibroids on transvaginal ultrasound [TVS], maximum fibroid diameter [MDF], uterine volume [UV], and FIGO-based leiomyoma subclassification [LS]). Additional variables included surgical approach (SA), postoperative fibroid residue, postoperative infection (POI) , oral contraceptive use, and postoperative pregnancy or childbirth (POP or POC). POP or POC was coded as a binary variable: 1 for occurrence, 0 for none. The BMI was categorized as follows: underweight ( 27.9 kg/m 2 ). Uterine volume was calculated using the ellipsoid formula 9 . Fibroid types were classified using the FIGO system: submucosal (FIGO 0–2) vs. others (FIGO 3–8). Residual fibroids were defined as any of the following: (1) fewer fibroids removed than indicated by preoperative ultrasound; (2) gynecologist-confirmed incomplete removal; (3) fibroids detected during the first postoperative ultrasound (within 3 months). i. Age between 18 and 44 years; ii. Hospitalization for uterine fibroid treatment; iii. No history of myomectomy. Age between 18 and 44 years; Hospitalization for uterine fibroid treatment; No history of myomectomy. i. Inability to complete follow-up via phone or refusal to disclose outcomes; ii. Known immune disorders (e.g., rheumatoid arthritis, lupus); iii. Absence of routine ultrasound follow-up (every 1–2 years); iv. pathology showing non-leiomyoma lesions; v. Hysterectomy during follow-up for other gynecologic reasons; vi. Congenital uterine anomalies; vii. Diagnosis of malignant tumors. Inability to complete follow-up via phone or refusal to disclose outcomes; Known immune disorders (e.g., rheumatoid arthritis, lupus); Absence of routine ultrasound follow-up (every 1–2 years); pathology showing non-leiomyoma lesions; Hysterectomy during follow-up for other gynecologic reasons; Congenital uterine anomalies; Diagnosis of malignant tumors. The full inclusion process is shown in the flowchart (Fig.  1 ). Fig. 1 The flowchart illustrates the screening process from the total number of patients to the final study sample. The flowchart illustrates the screening process from the total number of patients to the final study sample. Variable selection was performed using LASSO regression to identify key predictors of recurrence. Significant variables were then entered into a multivariate logistic regression model to determine independent risk factors. A nomogram prediction model was constructed using the “nomogram” function in the R software (v.4.4.0). The receiver operating characteristic (ROC) curve was plotted using the “roc” function, and the area under the curve (AUC) was calculated with the “auc” function to assess the discriminative ability of the model. The calibration curve was drawn using the “calibrate” function to evaluate the model’s calibration. The Hosmer–Lemeshow goodness-of-fit test was conducted using the “hoslem.test” function. The clinical decision curve was plotted using the “decision_curve” function to assess the clinical utility of the model. The model’s performance was evaluated based on sensitivity, specificity, positive predictive value, and negative predictive value. The CBCgrps package was used for statistical analysis. Continuous variables were tested for normality using the Shapiro–Wilk test. Data that followed a normal distribution were presented as mean ± standard deviation, and group comparisons were conducted using the t-test. For data that did not follow a normal distribution, the median (interquartile range) was used, and group comparisons were performed using the Mann–Whitney U test. Categorical variables were expressed as frequencies (%), and group comparisons were carried out using the chi-square (χ 2 ) test. Statistical analyses were performed using R software (v.4.4.0). A significance level of α = 0.05 was set, and differences were considered statistically significant when p  < 0.05.

The clinical cases included were randomly divided into a training cohort and a validation cohort at a 7:3 ratio, with a total of 476 cases in the training cohort and 202 cases in the validation cohort. Chi-square tests were used to assess differences in various indicators between the two groups. The results are shown in Table 1 , and there were no statistically significant differences between the training and validation cohorts across all indicators ( p  > 0.05). Table 1 Baseline clinical characteristics of patients in the training and validation cohorts. Variables Total (n = 678) Validation (n = 202) Training (n = 476) p Group, n (%) 1  Validatio 402 (59) 120 (59) 282 (59)  Training 276 (41) 82 (41) 194 (41) Age, n (%) 0.434  18–30 and 41–44 315 (46) 99 (49) 216 (45)  31–40 363 (54) 103 (51) 260 (55) BMI, n (%) 1  > 25 kg/m2 97 (14) 29 (14) 68 (14)  ≤ 25 kg/m2 581 (86) 173 (86) 408 (86) LS, n (%) 0.243  Others 533 (79) 165 (82) 368 (77)  Submucosal 145 (21) 37 (18) 108 (23) UV, n (%) 0.539  > 1140cm3 108 (16) 29 (14) 79 (17)  ≤ 1140cm3 570 (84) 173 (86) 397 (83) MDF, n (%) 0.51  > 4 cm 514 (76) 157 (78) 357 (75)  ≤ 4 cm 164 (24) 45 (22) 119 (25) Postoperative GnRHα, n (%) 0.725  No 643 (95) 193 (96) 450 (95)  Yes 35 (5) 9 (4) 26 (5) SA, n (%) 0.109  HL 547 (81) 171 (85) 376 (79)  T 131 (19) 31 (15) 100 (21) Residue, n (%) 0.268  No 513 (76) 159 (79) 354 (74)  Yes 165 (24) 43 (21) 122 (26) Endometriosis, n (%) 0.825  No 627 (92) 188 (93) 439 (92)  Yes 51 (8) 14 (7) 37 (8) AD or AM, n (%) 0.948  No 640 (94) 190 (94) 450 (95)  Yes 38 (6) 12 (6) 26 (5) ABM, n (%) 0.165  No 613 (90) 188 (93) 425 (89)  Yes 65 (10) 14 (7) 51 (11) POP or POC, n (%) 0.836  Yes 303 (45) 92 (46) 211 (44)  No 375 (55) 110 (54) 265 (56) OC, n (%) 0.395  No 657 (97) 198 (98) 459 (96)  Yes 21 (3) 4 (2) 17 (4) POI, n (%) 1  No 644 (95) 192 (95) 452 (95)  Yes 34 (5) 10 (5) 24 (5) Parturition, n (%) 0.729  Yes 610 (90) 180 (89) 430 (90)  No 68 (10) 22 (11) 46 (10) FH, n (%) 0.921  No 460 (68) 136 (67) 324 (68)  Yes 218 (32) 66 (33) 152 (32) TVS, n (%) 0.61  1 364 (54) 106 (52) 258 (54)  2 103 (15) 28 (14) 75 (16)  3–5 127 (19) 38 (19) 89 (19)  > 5 84 (12) 30 (15) 54 (11) Data are presented as n (%). No statistically significant differences were observed between the groups ( p  > 0.05). Baseline clinical characteristics of patients in the training and validation cohorts. Data are presented as n (%). No statistically significant differences were observed between the groups ( p  > 0.05). In the training cohort, LASSO regression was applied to identify predictors associated with uterine fibroid recurrence based on clinical variables. The coefficient path plot (Fig.  2 A) illustrates how each variable’s regression coefficient changes with the regularization parameter λ. As λ increases, coefficients for less important variables shrink toward zero, leaving only the most relevant predictors. The optimal λ was determined using tenfold cross-validation (Fig.  2 B), with the λ corresponding to the minimum mean squared error (left vertical dashed line) selected to reduce overfitting and enhance model robustness. LASSO regression identified six key variables: fibroid subtype (non-submucosal types), MDF, Residue, POP or POC, FH and TVS. Fig. 2 LASSO Regression for Variable Selection. ( A ) Coefficient path plot; ( B ) cross-validation curve. The vertical dashed line on the left represents the λ value associated with the minimum mean squared error. LASSO Regression for Variable Selection. ( A ) Coefficient path plot; ( B ) cross-validation curve. The vertical dashed line on the left represents the λ value associated with the minimum mean squared error. These six variables were subsequently entered into a multivariate binary logistic regression model. As shown in Table 2 , submucosal leiomyoma was identified as an independent protective factor (OR = 0.381, P  = 0.015). In contrast, postoperative residue (OR = 10.746, P  < 0.001), POP or POC (OR = 4.121, P  < 0.001), and FH (OR = 2.045, P  = 0.003) were significant independent risk factors for recurrence. While a fibroid diameter ≤ 4 cm appeared to confer a protective effect (OR = 0.817), this did not reach statistical significance ( P  = 0.571). Similarly, the number of fibroids on TVS was not statistically significant ( P  = 0.129), although a trend toward increased risk with higher counts was observed. Table 2 Multivariate logistic regression analysis for predicting postoperative uterine fibroid recurrence in the training cohort. Variable B SE Wald OR with CI P (Intercept) − 2.190 0.306 51.169 0.112(0.06–0.201)  < 0.001 LS − 0.964 0.398 5.885 0.381(0.171–0.815) 0.015 MDF ≤ 4 cm − 0.202 0.357 0.321 0.817(0.403–1.642) 0.571 Residue 2.375 0.333 50.756 10.746(5.696–21.116)  < 0.001 POP or POC 1.416 0.246 33.012 4.121(2.565–6.759)  < 0.001 FH 0.716 0.244 8.577 2.045(1.269–3.314) 0.003 TVS 0.188 0.124 2.309 1.207(0.946–1.538) 0.129 Multivariate logistic regression analysis for predicting postoperative uterine fibroid recurrence in the training cohort. Based on the multivariate logistic regression results, a nomogram was constructed to predict the risk of fibroid recurrence following myomectomy (Fig.  3 ). Each of the six predictors included in the model is assigned a point value using a corresponding scale. The total score, obtained by summing the individual scores, yields an estimated probability of recurrence. To demonstrate the application of the nomogram, we selected a representative patient from the study cohort. This patient had a maximum fibroid diameter > 4 cm, a non-submucosal fibroid subtype, postoperative residual fibroids, and four fibroids identified on transvaginal ultrasound (TVS score = 2). The total score calculated was 280, corresponding to an estimated recurrence probability of 87.8%. Fig. 3 Nomogram for predicting postoperative uterine fibroid recurrence. The model includes six predictors: MDF, LS, POP or POC, Residue, FH, and TVS. Each predictor is assigned a point score, and the total score corresponds to an estimated recurrence probability. Red dashed lines illustrate an example patient. Higher total points indicate a higher probability of fibroid recurrence. Nomogram for predicting postoperative uterine fibroid recurrence. The model includes six predictors: MDF, LS, POP or POC, Residue, FH, and TVS. Each predictor is assigned a point score, and the total score corresponds to an estimated recurrence probability. Red dashed lines illustrate an example patient. Higher total points indicate a higher probability of fibroid recurrence. To comprehensively evaluate the model’s performance, ROC curves were generated for both the training and validation cohorts based on predicted probabilities and actual outcomes. In addition to the AUC, key classification metrics—including sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and accuracy—were calculated to provide a more complete assessment of predictive performance (Table 3 ). Table 3 Performance metrics of the predictive model in the training and validation cohorts. Parameter Train Test Sensitivity 0.686 (0.611–0.747) 0.622 (0.516–0.727) Specificity 0.840 (0.794–0.881) 0.833 (0.756–0.892) PPV 0.747 (0.679–0.807) 0.718 (0.609–0.812) NPV 0.795 (0.745–0.838) 0.763 (0.682–0.830) Accuracy 0.777 (0.733–0.811) 0.748 (0.678–0.797) Performance metrics of the predictive model in the training and validation cohorts. In the training cohort, the model yielded an AUC of 0.834 (95% CI: 0.796–0.873) (Fig.  4 A), indicating excellent discriminatory ability. In the validation cohort, the AUC was 0.799 (95% CI: 0.737–0.861) (Fig.  4 B), demonstrating good generalizability. The C-index derived from ten-fold cross-validation in the training cohort was 0.795 (95% CI: 0.673–0.918), further supporting the model’s robustness and discriminative capability. Fig. 4 ROC curves for the logistic prediction model. ( A ) Training cohort (n = 476, AUC = 0.834); ( B ) Validation cohort (n = 202, AUC = 0.799). ROC curves for the logistic prediction model. ( A ) Training cohort (n = 476, AUC = 0.834); ( B ) Validation cohort (n = 202, AUC = 0.799). Calibration analysis further demonstrated strong consistency between predicted and observed recurrence probabilities. In the training cohort, the calibration curve (Fig.  5 A) closely aligned with the ideal reference line, with a mean absolute error of 0.019. In the validation cohort (Fig.  5 B), the calibration curve also showed good agreement, with a mean absolute error of 0.035. These findings suggest that the model exhibits high calibration quality and reliable performance across different datasets. Fig. 5 Calibration curves for the logistic prediction model. The red curve indicates the apparent predicted probabilities, the blue curve shows the bootstrap-corrected estimates, and the dashed line represents perfect calibration. ( A ) Training cohort; ( B ) Validation cohort. Calibration curves for the logistic prediction model. The red curve indicates the apparent predicted probabilities, the blue curve shows the bootstrap-corrected estimates, and the dashed line represents perfect calibration. ( A ) Training cohort; ( B ) Validation cohort. The Hosmer–Lemeshow goodness-of-fit test (χ 2  = 5.362, P  = 0.616) indicated no significant deviation between the model’s predicted probabilities and observed outcomes ( P  > 0.05), supporting its good calibration and overall fit. To further assess clinical applicability, DCA was performed for both the training and validation cohorts (Fig.  6 ). The DCA curves demonstrated that the prediction model yielded a consistently greater net benefit than the “treat all” or “treat none” strategies across a broad spectrum of threshold probabilities, particularly between 0.1 and 0.5. This threshold range reflects clinically relevant scenarios in which decisions about postoperative surveillance or preventive treatment may be considered. The red curve (training cohort) and blue curve (validation cohort) both lie above the gray line (“All”) and black line (“None”) across this range, suggesting that the model can effectively stratify patients according to their recurrence risk. This enables clinicians to prioritize intervention for high-risk individuals while avoiding unnecessary treatment in low-risk cases. Therefore, the model not only exhibits strong predictive performance but also has the potential to optimise, improving patient outcomes, and enhancing healthcare resource allocation. Fig. 6 Decision curve analysis for the logistic prediction model. The red line represents the training cohort, and the blue line represents the validation cohort. The gray line (All) assumes all patients receive treatment, while the black line (None) assumes no patient receives treatment. Decision curve analysis for the logistic prediction model. The red line represents the training cohort, and the blue line represents the validation cohort. The gray line (All) assumes all patients receive treatment, while the black line (None) assumes no patient receives treatment.

In this cohort study of 678 post-myomectomy patients, we developed and internally validated a predictive nomogram model using LASSO regression to quantify uterine fibroid recurrence risk. Key predictors—including fibroid subtype, maximum diameter, residue, childbirth status, family history, and fibroid count—were integrated into a clinically accessible framework. While these predictors are individually well known, their combined use in a validated, visual prediction model offers a practical tool for personalized risk assessment and postoperative planning. Limitations such as the exclusion of genetic markers and lack of external validation will be addressed in future studies. Nevertheless, this model represents a significant step toward individualized care for patients undergoing fibroid surgery and provides a foundation for precision follow-up and intervention strategies.

Uterine fibroid recurrence after myomectomy remains a significant clinical issue, with recurrence rates ranging from 11 to 50% 10 . These recurrences often lead to menstrual irregularities, chronic pelvic pain, and bulk-related symptoms, significantly impairing quality of life 11 , 12 . In severe cases, fibroid recurrence can compromise fertility, posing long-term implications for reproductive health 13 , 14 . While recurrence risk is often estimated through clinical judgment and patient history, these methods lack standardization and are subject to variability 15 . In this context, our study aimed to develop a reproducible and clinically interpretable prediction model using a nomogram framework. Rather than identifying novel predictors, we integrated well-established clinical variables—including fibroid number, size, subtype, postoperative residue, childbirth history, and family history—into a unified model. Using LASSO for variable selection and logistic regression for model construction, we transformed conventional risk factors into a comprehensive, individualized prediction tool. This methodological innovation enhances clinical decision-making by shifting from subjective estimates to quantified recurrence risk assessment. Importantly, we addressed a key limitation highlighted by Keizer et al. (2024), who identified gaps in existing models, such as poor clinical applicability and lack of internal validation 16 . Our model responds directly to these critiques by incorporating accessible clinical variables and undergoing rigorous internal validation. It offers clinicians a reliable tool for stratifying recurrence risk post-myomectomy and tailoring follow-up strategies accordingly. To enhance clinical applicability, this nomogram-based approach can be applied at multiple postoperative time points using updated clinical data. By manually recording and plotting the predicted recurrence risks, clinicians can monitor individual risk trends over time. Although an interactive web-based tool has not yet been developed, we intend to advance its construction in future studies to further improve the model’s usability and integration into clinical practice. Among the predictive variables, fibroid characteristics—especially size, type, and number—play central roles. Submucosal fibroids were associated with a lower recurrence risk, likely due to more complete surgical excision. In contrast, larger fibroids and multiple nodules may complicate complete removal due to irregular distribution and enhanced angiogenesis. These findings align with the hormonal theories proposed by Narine et al. 17 , who highlighted the role of endocrine regulation and postoperative pregnancy in mitigating recurrence through hormonal suppression of residual fibroid growth. Family history also emerged as an independent risk factor, likely reflecting underlying genetic predispositions. This supports findings by Langton et al. 18 , who associated recurrence with hereditary influences, reinforcing the importance of including family history in risk stratification. These observations pave the way for future research into genetic markers, such as fumarate hydratase (FH) mutations, which are implicated in hereditary leiomyomatosis and renal cell cancer (HLRCC) syndrome. We also emphasize the importance of surgical approach, as technique-related differences may affect residual burden and recurrence rates. Minimally invasive techniques, while offering faster recovery, may be limited in tactile feedback and field of view, potentially leaving undetected fibroids. Yasushi et al. 19 reported recurrence rates exceeding 76.2% at 8 years post-laparoscopic myomectomy, significantly higher than those following open surgery. Nonetheless, the primary goal of myomectomy—especially in reproductive-aged women—is to alleviate symptoms while preserving uterine integrity, not necessarily to achieve complete fibroid eradication. Therefore, early recurrence should be interpreted within the context of reproductive goals. A recurrence-free interval of 3–5 years may suffice for conception, and overly aggressive surgery may compromise the myometrium. A thorough pre-operative assessment of fibroid location and number, combined with appropriate surgical modality selection, is essential to optimize outcomes. For example, hysteroscopic resection of submucosal fibroids has been associated with recurrence rates as low as 0.9% at 2 years 20 . Our study is not without limitations. As a single-center, retrospective analysis, it is susceptible to selection bias and lacks ethnic diversity. Additionally, lifestyle and genetic risk factors, including FH mutations, were not included in the model due to data constraints. We also acknowledge the absence of standardized perioperative hormone measurements, which may affect recurrence dynamics. Moreover, due to the limited number of robotic-assisted cases and short follow-up, these were excluded but will be analyzed in future expansions. To address these limitations, we conducted statistical controls and plan to extend the study through multicenter, prospective cohorts, incorporating radiomics and genomic markers for enhanced generalizability. A prospective biobank is being considered to support integration of genetic data in future analyses.

Although uterine fibroids are benign tumors, their high postoperative recurrence rate poses a significant clinical challenge 1 . Approximately 15–30% of patients experience recurrence within 1 to 3 years, with rates approaching 50% by five years 2 . Several studies have indicated that age, hormonal levels, reproductive history, and lifestyle factors may influence the development and recurrence of uterine fibroids 3 , but existing research has produced inconsistent conclusions regarding the specific effects of these factors 4 . Moreover, current predictive models have considerable limitations in assessing individual recurrence risk 5 , failing to provide sufficient support for clinical decision-making. The increasing availability of electronic medical records and hospital databases provides new opportunities for developing more accurate and individualized prediction tools 6 , 7 . Nonetheless, most existing models inadequately integrate key predictors and fail to achieve reliable performance 8 . To address these gaps, we aimed to develop a novel recurrence risk model using LASSO regression, an advanced variable selection technique. Our goal was to improve predictive accuracy and support clinical decision-making by constructing a practical, data-driven nomogram for patients undergoing myomectomy.