Diagnostic accuracy of preoperative alpha-fetoprotein as an ovarian tumor marker in children and adolescents: not as good as we thought?

Purpose: To evaluate the diagnostic accuracy of preoperative serum alpha-fetoprotein (AFP) levels in predicting malignancy risk in children and adolescents presenting with ovarian neoplasms.
Methods: In 110 girls aged 18 and below diagnosed with ovarian neoplasms, we retrospectively correlated preoperative serum AFP levels with histological diagnosis of germ cell tumor or immature teratoma (GCT/IT) versus non-GCT/IT, and benign versus non-benign. We determined area under receiver-operating characteristic curves (AUC), sensitivity, specificity, and likelihood ratios.
Results: Twenty patients (18.2 %) had non-benign ovarian neoplasms, of which 12 had GCT/IT (10.9 %). In diagnosing GCT/IT versus non-GCT/IT, specificity of preoperative serum AFP was 87.8 %, sensitivity 66.7 %, and AUC 0.853. Excluding infants to remove the effects of increased variance in AFP in this group, specificity improved (92.0 %), but not sensitivity (66.7 %); AUC was 0.926. Increasing AFP cutoff to two times upper normal limit improved specificity (94.9 %), but not sensitivity (66.7 %). For benign versus non-benign tumors, AFP specificity was only 88.9 % and sensitivity 50.0 %.
Conclusion: The diagnostic accuracy of preoperative serum AFP for detecting GCT/IT in girls was limited by poor sensitivity and positive predictive value. Excluding infants and raising cutoff levels improved specificity marginally. Clinicians should be aware of these limitations when using AFP in the preoperative evaluation of childhood ovarian neoplasms.