Development and temporal validation of a nomogram for estimating the risk of malignancy in surgically managed adnexal masses after hysterectomy.

BACKGROUND

Adnexal masses discovered after hysterectomy are common, but the proportion that is malignant is low, and existing risk-stratification systems such as O-RADS were not developed or validated in this population. We aimed to develop and temporally validate a nomogram estimating the probability of malignancy among patients who have already been selected for surgical management of a post-hysterectomy adnexal mass.

METHODS

We retrospectively studied consecutive patients who underwent re-operation for an adnexal mass after prior benign hysterectomy at Beijing Obstetrics and Gynecology Hospital between January 2010 and October 2025. Patients operated from January 2010 to September 2023 formed the development cohort (n = 233; 50 malignant, 183 benign); those operated from October 2023 to October 2025 formed the temporal validation cohort (n = 48; 20 malignant, 28 benign). Fourteen candidate predictors were screened by LASSO regression with 10-fold cross-validation. Final coefficients were obtained from unpenalised logistic regression on the selected predictors; the intercept of the seven-variable model was recalibrated on the development cohort because the originally reported intercept was inconsistent with the scaling of the continuous predictors. Performance was assessed by discrimination, calibration (intercept, slope, Brier score) and decision-curve analysis, with 1,000 bootstrap resamples for optimism correction.

RESULTS

The seven-variable nomogram comprises age (per 10 years), BMI (per 10 kg/m2), tumour size (ordinal), cystic components, enriched blood flow/enhancement, elevated CA125 (≥70 U/mL) and cyst duration (per 5 years). In the development cohort the AUC was 0.959 (95% CI 0.922-0.990) for the seven-variable model and 0.937 (0.885-0.979) for a simplified three-variable model (cystic components, enhancement, elevated CA125); the difference favoured the full model (ΔAUC -0.023, 95% CI -0.045 to -0.005; DeLong p = 0.021). In the temporal validation cohort the AUC was 0.905 (0.791-0.985) and 0.929 (0.846-0.986), respectively (ΔAUC +0.022, 95% CI -0.040 to 0.097; p = 0.517). Calibration was adequate in development (intercept 0.00, slope 1.02) and showed no statistically significant miscalibration in validation (intercept 0.29, 95% CI -0.71 to 1.28; slope 0.71, 95% CI 0.34 to 1.08). At the pre-specified 0.5 threshold the full model missed 8 of 50 malignancies in development and 6 of 20 in validation.

CONCLUSION

We present a parsimonious nomogram that estimates malignancy risk in post-hysterectomy adnexal masses among patients already selected for surgery, with internal and temporal validation. Because all participants underwent surgery, the model cannot be used to justify non-operative management; independent external and prospective validation in unselected cohorts is required before any clinical implementation.