ADNEX Model Predicts Risk of Malignancy and Stratifies Risk Across Tumor Subtypes in Women With Adnexal Mass

EBM Focus - Volume 9, Issue 42

Read the complete EBM Focus and earn CME credit.

Reference: BMJ 2014 Oct 15;349:g5920 (level 1 [likely reliable] evidence)

All 5,909 women included in the study were part of the International Ovarian Tumour Analysis (IOTA) study group, were assessed by transvaginal ultrasound, and later had surgical intervention. The Assessment of Different NEoplasias in the adneXa (ADNEX) model was developed from a cohort of 3,506 women and includes 9 criteria: age, serum CA-125 level, type of center (oncology center vs. other hospital), maximum lesion diameter, solid tissue proportion, presence of > 10 cyst locules, papillary projection number, presence of acoustic shadows, and presence of ascites. In the validation cohort of 2,403 women, the ADNEX model had excellent performance for discrimination of benign and malignant tumors (c-statistic 0.94). Using 3% as a cutoff for risk of malignancy, the ADNEX model had 98.9% sensitivity and 46.6% specificity. The positive and negative likelihood ratios were 1.9 and 0.02, respectively. Based on the negative likelihood ratio, the risk of malignancy with a negative result is very low. In addition to classifying the risk of a tumor being benign vs. malignant, the ADNEX model also stratifies the risk of malignancy into 4 subtypes (borderline tumor, stage I invasive, stage II-IV invasive, and secondary metastatic), all of which had good discrimination (c-statistic 0.75-0.96 in combined data set) and calibration.

The ADNEX model improves the classification of tumor type in women with adnexal masses. The National Institutes for Health and Care Excellence (NICE 2011 Apr: CG122) currently recommends using the risk of malignancy index (RMI) for assessing tumor risk, however in the IOTA data set the RMI had lower predictive performance than the ADNEX model (c-statistic 0.88). The ADNEX model also offers additional benefit by classifying risk of malignancy into several subtypes, which may inform patient triage as well as care decisions and management. The model is available as a free online calculator or can be purchased as an app for android mobile phones (iPhone app is not yet available).

For more information, see the Ovarian cancer and Adnexal mass topics in DynaMed.