Development and validation of a predictive nomogram for pediatric ovarian torsion: a retrospective cohort study

Introduction

Pediatric ovarian torsion is an uncommon yet critical surgical emergency. Delayed diagnosis may result in ovarian ischemia and necrosis, and may increase the risk of unnecessary oophorectomy, thereby adversely affecting long-term reproductive and endocrine function (1,2). Therefore, timely and accurate diagnosis followed by prompt surgical intervention is essential to preserve ovarian viability and future fertility in affected children (3).

However, the clinical diagnosis of pediatric ovarian torsion is especially challenging due to non-specific symptoms, communication barriers in infants and young children, and limited cooperation during physical examination and imaging. To date, studies exploring predictive factors for pediatric ovarian torsion remain limited (4-8). Although previous studies have highlighted the diagnostic value of sonographic findings and the promise of the neutrophil-to-lymphocyte ratio (NLR) as an inflammatory marker, most studies have been constrained by small sample sizes and a lack of validation, thereby limiting the robustness and generalizability of their predictive models.

To address the urgent need for improved early diagnostic capabilities, this study focuses on developing and validating a robust risk prediction model by integrating multidimensional clinical, laboratory, and sonographic data. We adopted a nomogram-based approach to visualize the predictive model and facilitate individualized risk quantification in clinical practice (9). This user-friendly tool is designed to support clinicians in the rapid bedside identification of high-risk patients, thereby guiding more precise and individualized clinical management. We present this article in accordance with the TRIPOD reporting checklist (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-1-0131/rc).

Methods

Study population

This retrospective study involved pediatric patients who underwent emergency surgery for suspected ovarian torsion at the Department of Surgical Oncology, Children’s Hospital, Zhejiang University School of Medicine, between October 20, 2017 and October 20, 2025. The inclusion criteria were as follows: (I) pediatric patients (aged <18 years) and (II) those who underwent emergency surgery based on a presumptive preoperative diagnosis of ovarian torsion. The exclusion criterion was patients with incomplete clinical or imaging data.

Ethics approval statement

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Children’s Hospital, Zhejiang University School of Medicine (No. 2025-IRB-0010-P-01). Given the retrospective design and the use of de-identified patient data, the requirement for informed consent was waived by the Institutional Review Board.

Data collection

Data was retrospectively collected from electronic medical records. Collected variables included: (I) demographics and clinical details: age, height, weight, body mass index (BMI), menarcheal status, history of abdominal or pelvic surgery, symptoms duration (hours), presenting symptoms (abdominal pain, vomiting, fever), and physical examination findings (abdominal tenderness, rebound tenderness). (II) Laboratory parameters: white blood cell (WBC) count, absolute neutrophil and lymphocyte counts, and platelet (PLT) count, hemoglobin (Hb) level, and high-sensitivity C-reactive protein (CRP). The NLR was calculated by dividing the absolute neutrophil count by the absolute lymphocyte count. (III) Sonographic features: maximum ovarian diameter, presence of an ovarian mass, whirlpool sign, and pelvic free fluid. (IV) Surgical and histopathological outcomes: intraoperative findings serving as the reference standard for diagnosis. Key surgical variables included intraoperative confirmation of ovarian torsion, lesion laterality, torsion degree, and performance of oophorectomy. Histopathological results were classified into four categories: solid tumors (including mature/immature teratomas, cystadenomas, and sex cord tumors), cysts, hematomas, and no underlying lesion.

Handling of missing data

Patients with missing ultrasound data (n=10) were excluded from the analysis. Missing height data (n=14) were handled using single imputation, whereby values were replaced with the age-specific median height based on the Health Industry Standard of the People’s Republic of China (10,11).

Statistical analysis

Statistical analyses were conducted using R software (version 4.5.1) within the RStudio integrated development environment (version 2025.5.1.513). Key packages utilized included rms for nomogram construction and calibration, pROC for receiver operating characteristic (ROC) analysis, and rmda for decision curve analysis (DCA). The normality of continuous variables was assessed using the Shapiro-Wilk test. Normally distributed data are presented as mean ± standard deviation (SD) and were compared through independent t-tests. Non-normally distributed data are presented as median and interquartile range (IQR) and were compared using the Mann-Whitney U test. Categorical data are reported as counts and percentages (n, %) and were compared using the Chi-squared test or Fisher’s exact test, as appropriate. A two-tailed P value <0.05 was considered statistically significant.

Model development and internal validation

The cohort was randomly partitioned into a training set (70%) for model development and a validation set (30%) for internal split-sample validation using a computerized random seed. Univariate logistic regression analyses were first conducted on the training set to identify potential predictors of ovarian torsion. For variables exhibiting quasi-complete separation, such as rebound tenderness, Firth’s penalized likelihood logistic regression was utilized (12). Variables with a P value <0.05 in the univariate analysis were subsequently incorporated into a multivariate logistic regression model. This two-step approach was adopted to build a parsimonious model by first identifying all potentially relevant factors and then selecting only those that remained as independent predictors, thereby minimizing overfitting and enhancing the model’s clinical applicability. Independent predictors identified in the multivariate analysis (P<0.05) were used to construct the final predictive nomogram. Associations were quantified using the odds ratio (OR) with a 95% confidence interval (CI). Nomogram points for each predictor were assigned by proportionally scaling their regression coefficients (β values), such that the variable with the largest effect size was assigned a maximum of 100 points.

Assessment of model performance

The performance of the nomogram was evaluated in both the training and validation sets across three domains: discrimination, calibration and clinical utility. Discrimination, reflecting the ability of the model to distinguish between patients with and without ovarian torsion, was assessed using the area under the receiver operating characteristic curve (AUC) with 95% CI. Calibration, representing the agreement between the nomogram-predicted probabilities and observed outcomes, was evaluated using calibration plots and quantified using the Brier score (13). In addition, the Spiegelhalter Z-test was used to test for significant deviation from perfect calibration. Clinical utility was assessed using DCA, which quantified the net benefit of the nomogram across a range of clinically relevant threshold probabilities (14).

Results

Baseline characteristics of the study cohort

A total of 201 pediatric patients were included in the study and randomly allocated to a training set of 140 and a validation set of 61. Baseline demographic and clinical characteristics of the two sets are summarized in Table 1. The overall median age was 10.3 (IQR, 7.6–12.3) years, and the median duration of symptoms was 24 (IQR, 18.0–72.0) hours. Abdominal pain was the most common presenting symptom, reported in 196 patients (97.5%). In the remaining five non-verbal patients, symptoms were documented as inconsolable crying accompanied by vomiting. The median NLR was 3.7 (IQR, 2.3–6.6), and the median maximum ovarian diameter was 5.9 (IQR, 4.8–7.8) cm. An underlying ovarian mass was identified in 180 patients (89.6%), and the whirlpool sign was visualized in 63 patients (31.3%). No statistically significant differences were observed between the training and validation sets with respect to baseline characteristics (all P>0.05), indicating good comparability between the two groups.

Surgical and histopathological outcomes

The surgical and histopathological findings, stratified by the final diagnosis of ovarian torsion, are summarized in Table 2. Ovarian torsion was intraoperatively confirmed in 153 of 201 patients (76.1%). Among these 153 confirmed cases, torsion occurred more frequently on the right side (94/153, 61.4%), and a median torsion degree of 720° (IQR, 360°–720°). Oophorectomy was exclusively performed in patients with confirmed torsion; however, it was required in only 9 cases (5.9%), indicating a high rate of ovarian preservation.

The distribution of histopathological diagnoses differed significantly between the torsion and non-torsion groups (P<0.001). Solid tumors were the most common underlying pathology in the torsion group, identified in 94 patients (61.4%), compared with 18 patients (37.5%) in the non-torsion group, with mature teratoma being the predominant subtype. In contrast, hematomas were substantially more prevalent in the non-torsion group than in the torsion group (20.8% vs. 2.0%).

Univariate logistic regression analysis

The results of the univariate logistic regression analysis in the training set are presented in Table 3. Abdominal pain, present in 196 of 201 patients (97.5%), was excluded from the analysis due to its limited discriminative value. Rebound tenderness, analyzed using Firth’s regression, was not significantly associated with ovarian torsion (P=0.22). Among the included variables, vomiting was identified as a significant risk factor (OR =4.84, P<0.001), whereas post-menarcheal status (P<0.001), older age (P<0.001), and longer symptom duration (P=0.04) were associated with a significantly lower risk of torsion.

Regarding laboratory parameters, both elevated WBC count (P=0.006) and increased NLR (P<0.001) were significantly associated with ovarian torsion. Sonographic findings demonstrated strong predictive value, with the whirlpool sign showing the highest predictive value (OR =8.23, P<0.001). Additionally, increasing maximum ovarian diameter was significantly associated with torsion (OR =1.63 per cm, P<0.001). Variables with a P value <0.05 in univariate analyses were subsequently entered into multivariate analysis.

Multivariate analysis and nomogram construction

Variables demonstrating statistical significance in the univariate analysis were incorporated into a multivariate logistic regression model. The final model comprising four independent predictors of pediatric ovarian torsion: the whirlpool sign, maximum ovarian diameter, NLR, and age (Table 4). The whirlpool sign remained the strongest predictor (OR =7.09, P=0.009). Both increasing ovarian diameter (OR =1.75 per cm, P=0.001) and higher NLR (OR =1.23, P=0.02) were independently associated with an increased risk of torsion. Notably, age was independently and inversely associated with ovarian torsion; each additional year of age was associated with a 33% reduction in risk (OR =0.67, P=0.001). These four independent predictors were subsequently utilized to create a predictive nomogram for clinical application (Figure 1), generating a total point score that aligns with an individual’s predicted likelihood of ovarian torsion.

Figure 1 Nomogram for predicting the risk of pediatric ovarian torsion. The model incorporates four predictors. A total score is calculated by summing the points assigned to each predictor, which corresponds to a specific predicted probability of torsion on the final axis. Whirlpool sign: “0” indicates absence of the whirlpool sign, and “1” indicates presence of the whirlpool sign. NLR, neutrophil-to-lymphocyte ratio.

Predictive performance of the nomogram

The predictive performance of the nomogram was evaluated in terms of discrimination, calibration, and clinical utility. The model exhibited excellent discrimination, with an AUC of 0.915 (95% CI: 0.865–0.964) in the training set and maintained robust performance in the validation set, achieving an AUC of 0.853 (95% CI: 0.711–0.994) (Figure 2).

Figure 2 ROC curves of the predictive nomogram. The model’s performance was assessed in both the training set and the validation set. The AUC was 0.915 (95% CI: 0.865–0.964) in the training set and 0.853 (95% CI: 0.711–0.994) in the validation set, demonstrating excellent discrimination in both cohorts. AUC, area under the curve; CI, confidence interval; ROC, receiver operating characteristic.

Calibration analysis demonstrated good agreement between predicted and observed probabilities in both cohorts (Figure 3). The Brier scores were low in both the training set (0.099) and the validation set (0.102), indicating good overall model accuracy. The Spiegelhalter Z-test further confirmed that the model was well-calibrated in the validation cohort, showing no significant deviation between predictions and actual outcomes (P=0.90).

Figure 3 Calibration curves of the nomogram for predicting ovarian torsion in the training and validation sets. The plot compares the predicted probability of ovarian torsion (x-axis) with the observed frequency (y-axis) for both the training and validation sets. The 45-degree dashed line represents perfect calibration. The close proximity of the model’s curves to the dashed line indicates a high degree of agreement between prediction and reality.

Clinical utility was assessed using DCA (Figure 4). Across a wide range of threshold probabilities, the nomogram consistently provided a greater net benefit than either treating all patients or none in both cohorts. For instance, at a threshold probability of 20%, the net benefit was 0.711 in the training set and 0.729 in the validation set, supporting the potential clinical usefulness of the model.

Figure 4 DCA for the predictive nomogram. The plot displays the net benefit (y-axis) at various threshold probabilities (x-axis). The DCA demonstrates that using the nomogram (red and blue line) to inform clinical decisions offers a greater net benefit than the strategies of treating all patients (gray line) or treating none (black line) across a wide range of threshold probabilities. DCA, decision curve analysis.

Discussion

This study developed and internally validated a nomogram for predicting pediatric ovarian torsion by integrating four clinically relevant predictors: age, NLR, the whirlpool sign, and maximum ovarian diameter. The model demonstrated strong discriminatory performance in distinguishing torsion from non-torsion cases. Furthermore, calibration analysis confirmed its predictive accuracy, while DCA affirmed its clinical utility by showing a clear net benefit. Collectively, these findings suggest that the proposed nomogram may serve as a practical, quantitative bedside tool to assist clinicians in the early identification of pediatric patients with suspected ovarian torsion.

Non-invasive ultrasonography is the cornerstone of diagnosis for pediatric ovarian torsion. In line with previous reports, our findings confirm that the whirlpool sign is the most powerful predictor (OR =7.09), reflecting direct visualization of a twisted vascular pedicle and thus possesses high specificity (15). However, the detectability of the whirlpool sign may be influenced by sonographer experience, ovarian position, or patient cooperation, and it is not identified in all cases of torsion. Consequently, the absence of a whirlpool sign does not exclude the diagnosis. In this context, a significantly enlarged ovary should maintain a high index of suspicion for ovarian torsion (16). The inclusion of maximum ovarian diameter (OR =1.75 per cm) as another strong, independent predictor is a key feature of our nomogram. However, several additional sonographic features, including the follicular ring sign, ovarian diameter ratio, abnormal ovarian position and abnormal Doppler blood flow (17,18), were not routinely assessed or recorded in our cohort. This was largely due to limited patient cooperation and the time-sensitive nature of emergency examinations, leading to variability in historical documentation. Consequently, these factors were not included in the analysis. This represents a limitation of our model, and future studies employing standardized pediatric ultrasound protocols are essential to systematically evaluate these parameters and further improve diagnostic performance.

While ovarian torsion can occur across the female lifespan, the present study specifically focused on the pediatric and adolescent population (<18 years). Within this cohort, age emerged as a significant independent inverse predictor (OR 0.67 per year), indicating a higher risk of torsion with decreasing age. This association is likely attributable to anatomical factors, including relatively elongated and more mobile adnexal ligaments in infants and prepubertal girls. Such features increase ovarian mobility and susceptibility to torsion, particularly in the presence of a physiological cyst or adnexal mass (19). The observed inverse relationship between age and torsion risk is consistent with clinical guidance outlined in the American College of Obstetricians and Gynecologists (ACOG) Committee Opinion on adolescents (20). However, it is essential to clarify that this age-related trend is specific to the pediatric population and should not be extrapolated to adult patients. Furthermore, younger children may have difficulty accurately localizing or verbalizing symptoms, which can contribute to diagnostic delays. Therefore, clinicians should maintain a high index of suspicion in this population. In cases of persistent acute abdominal pain where common differential diagnoses, such as appendicitis, have been reasonably excluded (21), prompt pelvic ultrasonography is strongly recommended.

The NLR, a readily available systemic inflammatory marker, reflects the ischemic-reperfusion injury and inflammatory cascade initiated by the torsion event. In this study, an elevated NLR was independently associated with an increased risk of ovarian torsion (OR =1.23), a finding consistent with previous studies (5). These results support the role of NLR as a useful adjunctive biomarker in the early diagnostic evaluation of pediatric patients with suspected torsion, particularly when imaging findings are equivocal.

The principal contribution of this study is the integration of diverse predictors into a comprehensive, user-friendly nomogram for individualized risk estimation (9). While previous clinical scores (7,22), provided valuable diagnostic frameworks, they often lacked objective laboratory or imaging parameters. Similarly, emerging single biomarkers like preoperative D-dimer, despite showing high sensitivity, lack the independent predictive power to replace imaging (23). These findings highlight the superior utility of our multi-marker approach. By synergistically combining a systemic inflammatory marker (NLR) with highly specific sonographic findings (whirlpool sign and ovarian diameter), our nomogram yields a more robust risk assessment than isolated scores or biomarkers. In the high-stakes pediatric emergency setting, this enables rapid, objective stratification: high scores expedite surgical intervention to maximize ovarian salvage, while low scores support safe observation, effectively reducing unnecessary invasive procedures.

Importantly, this nomogram is intended to serve as a decision-support tool rather than a definitive diagnostic test. Consequently, we do not advocate for a single, universal cutoff probability to mandate surgical intervention. The decision for exploratory laparoscopy must remain individualized, balancing the nomogram’s predicted risk against the overall clinical picture and the inherent risks of negative exploration versus missed torsion. The broad net benefit demonstrated in our DCA supports the model’s flexibility and utility across varying clinical thresholds.

Several limitations of this study should be acknowledged. First, the retrospective, single-center design may introduce selection or information biases, which could limit the generalizability of the findings. Although internal validation demonstrated robust model performance, external validation using prospective, multicenter cohorts is required to confirm its applicability across diverse clinical settings. Second, the study population was restricted to pediatric and adolescent patients, with an age range of 2 to 201 months (median 10.3 years). Therefore, the findings should not be extrapolated to the adult population. Third, while this cohort represents one of the larger single-center series of pediatric ovarian torsion, larger sample sizes would allow for the inclusion of additional candidate predictors and more precise risk estimation. Finally, surgical confirmation was used as the diagnostic reference standard; consequently, patients with suspected torsion who were managed non-operatively, such as those experiencing spontaneous detorsion, may not have been captured in the analysis.

Conclusions

In conclusion, this study developed and internally validated a predictive nomogram integrating age, NLR, the whirlpool sign, and maximum ovarian diameter for the assessment of pediatric ovarian torsion. The model exhibits good discrimination and calibration, indicating its accuracy and clinical value. This nomogram may serve as a practical, quantitative tool to support timely diagnosis and clinical decision-making, with the goal of improving ovarian preservation in pediatric patients.

Acknowledgments

We would like to thank Litao Zhou from the Statistics Office and Yichan Wan from the Department of Medical Records and Archives for their invaluable assistance with data retrieval for this study.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-1-0131/rc

Data Sharing Statement: Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-1-0131/dss

Peer Review File: Available at https://tp.amegroups.com/article/view/10.21037/tp-2026-1-0131/prf

Funding: This work was supported by a grant of “Pioneer” and “Leading Goose” R&D Program of Zhejiang Province (No. 2024C03181, to J.W.).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tp.amegroups.com/article/view/10.21037/tp-2026-1-0131/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Ethics Committee of Children’s Hospital, Zhejiang University School of Medicine (No. 2025-IRB-0010-P-01). Given the retrospective design and the use of de-identified patient data, the requirement for informed consent was waived by the Institutional Review Board.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Amir N, McCracken K. Pediatric ovarian torsion. Semin Pediatr Surg 2025;37:151546. [Crossref] [PubMed]
2. Tielli A, Scala A, Alison M, et al. Ovarian torsion: diagnosis, surgery, and fertility preservation in the pediatric population. Eur J Pediatr 2022;181:1405-11. [Crossref] [PubMed]
3. Dasgupta R, Renaud E, Goldin AB, et al. Ovarian torsion in pediatric and adolescent patients: A systematic review. J Pediatr Surg 2018;53:1387-91. [Crossref] [PubMed]
4. Boztas AE, Sencan E, Payza AD, et al. An Innovative Machine Learning-Based Algorithm for Diagnosing Pediatric Ovarian Torsion. J Pediatr Surg 2025;60:162417. [Crossref] [PubMed]
5. Delgado-Miguel C, Arredondo-Montero J, Moreno-Alfonso JC, et al. The Role of Neutrophyl-to-Lymphocyte Ratio as a Predictor of Ovarian Torsion in Children: Results of a Multicentric Study. Life (Basel) 2024;14:889. [Crossref] [PubMed]
6. Chen S, Gao Z, Qian Y, et al. Key clinical predictors in the diagnosis of ovarian torsion in children. J Pediatr (Rio J) 2024;100:399-405. [Crossref] [PubMed]
7. Meyer R, Meller N, Mohr-Sasson A, et al. A clinical prediction model for adnexal torsion in pediatric and adolescent population. J Pediatr Surg 2022;57:497-501. [Crossref] [PubMed]
8. Rey-Bellet Gasser C, Gehri M, Joseph JM, et al. Is It Ovarian Torsion? A Systematic Literature Review and Evaluation of Prediction Signs. Pediatr Emerg Care 2016;32:256-61.
9. Balachandran VP, Gonen M, Smith JJ, et al. Nomograms in oncology: more than meets the eye. Lancet Oncol 2015;16:e173-80. [Crossref] [PubMed]
10. National Health Commission of the People’s Republic of China. Standard for height level classification among children and adolescents aged 7~18 years. 2018. [accessed 19 November 2025]. Available online: https://www.nhc.gov.cn/wjw/pqt/201807/a5269d062966455ea55d90a02c690311.shtml
11. National Health Commission of the People’s Republic of China. Growth standard for children under 7 years of age. 2022. [accessed 19 November 2025]. Available online: https://www.nhc.gov.cn/wjw/c100311/202211/923e7646561d4b88b72da9097d4da4d5.shtml
12. Heinze G, Schemper M. A solution to the problem of separation in logistic regression. Stat Med 2002;21:2409-19. [Crossref] [PubMed]
13. Van Calster B, McLernon DJ, van Smeden M, et al. Calibration: the Achilles heel of predictive analytics. BMC Med 2019;17:230. [Crossref] [PubMed]
14. Vickers AJ, Elkin EB. Decision curve analysis: a novel method for evaluating prediction models. Med Decis Making 2006;26:565-74. [Crossref] [PubMed]
15. Valsky DV, Esh-Broder E, Cohen SM, et al. Added value of the gray-scale whirlpool sign in the diagnosis of adnexal torsion. Ultrasound Obstet Gynecol 2010;36:630-4. [Crossref] [PubMed]
16. Oltmann SC, Fischer A, Barber R, et al. Cannot exclude torsion--a 15-year review. J Pediatr Surg 2009;44:1212-6; discussion 1217. [Crossref] [PubMed]
17. Almalki YE, Basha MAA, Nada MG, et al. Sonographic signs of pediatric ovarian torsion: Multicenter study with surgical correlation. Eur J Radiol 2025;190:112258. [Crossref] [PubMed]
18. Hartman SJ, Prieto JM, Naheedy JH, et al. Ovarian volume ratio is a reliable predictor of ovarian torsion in girls without an adnexal mass. J Pediatr Surg 2021;56:180-2. [Crossref] [PubMed]
19. Sintim-Damoa A, Majmudar AS, Cohen HL, et al. Pediatric Ovarian Torsion: Spectrum of Imaging Findings. Radiographics 2017;37:1892-908. [Crossref] [PubMed]
20. Adnexal Torsion in Adolescents. ACOG Committee Opinion No, 783. Obstet Gynecol 2019;134:e56-63.
21. Buel KL, Wilcox J, Mingo PT. Acute Abdominal Pain in Children: Evaluation and Management. Am Fam Physician 2024;110:621-31.
22. Bolli P, Schädelin S, Holland-Cunz S, et al. Ovarian torsion in children: Development of a predictive score. Medicine (Baltimore) 2017;96:e8299. [Crossref] [PubMed]
23. Boscarelli A, Giangreco M, Valvo E, et al. The role of preoperative D-dimer levels in diagnosing adnexal torsion in children and adolescents: a prospective monocentric pilot analysis. Front Med (Lausanne) 2025;12:1706633. [Crossref] [PubMed]