Validity of predictions of residual retroperitoneal mass histology in nonseminomatous testicular cancer

• Published in: Journal of clinical oncology Vol. 16; no. 1; p. 269
• Main Authors: Steyerberg, E W, Gerl, A, Fossá, S D, Sleijfer, D T, de Wit, R, Kirkels, W J, Schmeller, N, Clemm, C, Habbema, J D, Keizer, H J
• Format: Journal Article
• Language: English
• Published: United States 01-01-1998
• Subjects: Antineoplastic Combined Chemotherapy Protocols - therapeutic use; Area Under Curve; Humans; Male; Models, Biological; Models, Statistical; Necrosis; Neoplasm, Residual; Reproducibility of Results; Retroperitoneal Neoplasms - drug therapy; Retroperitoneal Neoplasms - pathology; Retroperitoneal Neoplasms - secondary; Teratoma - drug therapy; Teratoma - pathology; Teratoma - secondary; Testicular Neoplasms - drug therapy; Testicular Neoplasms - pathology
• ISSN: 0732-183X
• DOI: 10.1200/JCO.1998.16.1.269

To validate predictions of the histology (necrosis, mature teratoma, or cancer) of residual retroperitoneal masses in patients treated with chemotherapy for metastatic nonseminomatous testicular germ cell tumor.
We studied 172 testicular cancer patients who underwent resection while tumor markers were normal. Predictive characteristics for the residual histology were registered, including the presence of teratoma elements in the primary tumor, the prechemotherapy level of tumor markers (alpha-fetaprotein [AFP], human chorionic gonadotropin [HCG], lactate dehydrogenase [LDH]), the size of the residual mass, and the percentage of shrinkage in mass diameter. We calculated the predicted probability of necrosis and the ratio of cancer and mature teratoma with previously published logistic regression formulas.
The distribution of the residual histology was necrosis in 77 (45%), mature teratoma in 72 (42%), and cancer in 23 (13%). Necrosis could be well distinguished from other tissue, with an area under the receiver operating characteristic (ROC) curve of 82%. No tumor was found in 15 patients with a predicted probability of necrosis over 90%. The predicted probabilities corresponded reliably with the observed probabilities (goodness-of-fit tests, P > .20), although a somewhat higher probability of necrosis was observed in patients treated with chemotherapy containing etoposide. Conversely, cancer could not reliably be predicted or adequately discriminated from mature teratoma.
The predicted probabilities of necrosis have adequate reliability and discriminative power. These predictions may validly support the decision-making process regarding the need and extent of retroperitoneal lymph node dissection.