Predicting the Need for Surgery in Tubo-Ovarian Abscess: A Clinical Evaluation and External Validation of Risk Scores

This is the authors' abstract. We don't add key points for this paper.

OBJECTIVE: To investigate clinical and imaging factors predictive of successful conservative management of tubo-ovarian abscesses (TOA), and to validate two existing predictive scores. DESIGN: Retrospective single-center study. SETTING: Academic medical center in Leipzig, Germany. PARTICIPANTS: A total of 166 women hospitalized with TOA between 2010 and 2022. INTERVENTIONS: Patients were grouped by treatment approach (conservative, surgical, or emergency surgery) and analyzed using descriptive statistics and logistic regression. Two previously published risk scores (Fouks and Yongue) were externally validated using calibration plots, area under the curve (AUC), and decision curve analysis. RESULTS: Of 118 patients requiring surgery, 67.8% underwent surgery during initial admission, while the remainder were re-admitted later during the disease course due to failure of conservative treatment. Factors associated with surgical intervention included older age, larger abscess diameter, microbiological proof of Gardnerella vaginalis infection, and abdominal guarding. In a multivariable regression model, G. vaginalis infection (OR 6.9, 95% confidence interval [CI] 1.6-49.6, p = .02), abdominal guarding (OR 3.1, 95% CI 1.1-9.5, p = .04), and recent abdominal/gynecological intervention (OR 5.6, 95% CI 1.2-40.5, p = .04) were independent predictors of the need for surgery. Both previously published risk scores demonstrated poor discriminatory ability (Fouks: AUC = 0.65; Yongue: AUC = 0.55), with no significant difference between them. CONCLUSION: Predictive factors for conservative treatment failure in TOA remain inconsistent. Currently available risk scores could not reliably predict the need for surgical intervention in our cohort, stressing the importance of external validation before clinical implementation. Further multi-center studies are needed to develop robust prediction models for clinical practice.