ADNEX risk prediction model for diagnosis of ovarian cancer: systematic review and meta-analysis of external...

Area of Science:

• Gynecologic Oncology
• Medical Diagnostics
• Machine Learning in Healthcare

Background:

• Ovarian cancer diagnosis relies on accurate risk stratification of adnexal masses.
• The Assessment of Different Neoplasias in the adnexa (ADNEX) model is a multivariable tool for diagnosing ovarian cancer.
• External validation is crucial to assess the generalizability and performance of predictive models.

Purpose of the Study:

• To systematically review and meta-analyze external validation studies of the ADNEX model.
• To evaluate the diagnostic performance of the ADNEX model in distinguishing benign from malignant ovarian tumors.
• To assess the clinical utility and generalizability of the ADNEX model across different settings.

Main Methods:

• Systematic review and meta-analysis of 47 external validation studies (17,007 tumors) published between 2014 and 2023.
• Data extraction and quality assessment using TRIPOD and PROBAST guidelines.
• Random effects meta-analysis of diagnostic performance metrics (AUC, sensitivity, specificity) and clinical utility.

Main Results:

• The ADNEX model demonstrated high diagnostic accuracy, with a summary AUC of 0.93 for distinguishing benign from malignant tumors, both with and without CA125.
• The model showed high estimated probabilities of clinical utility (91-95%) in new centers.
• Most validation studies (91%) were at high risk of bias, primarily due to issues with case selection and lack of calibration assessment.

Conclusions:

• The ADNEX model is a robust tool for differentiating benign and malignant adnexal masses across various international settings.
• The model's performance is consistent with or without the inclusion of CA125.
• Further research should focus on improving the reporting and assessment of model calibration in external validation studies.