Three models to predict axillary lymph node involvement in patients with breast cancer validated

Although models have clinical utility, usefulness may still be limited.

When tested, three models used to predict involvement of four or more axillary lymph nodes in patients with breast cancer with one to three metastatic sentinel lymph nodes were all validated: the Louisville score excluding method of detection, the Louisville score including method of detection and the Katz nomogram.

However, the Katz nomogram, developed by Katz et al, outperformed the two Louisville scores, developed by Chagpar et al.

Patients with breast cancer considering reconstruction surgery have to meet certain criteria before the procedure can be carried out, according to background information. This is because postmastectomy radiation therapy can lead to adverse effects once reconstruction is completed. The American Society of Clinical Oncology and the American Society for Therapeutic Radiology and Oncology both recommend postmastectomy radiation in patients who have four or more involved axillary lymph nodes.

Recently, three models were developed to predict if four or more axillary lymph nodes are involved, the two by Chagpar et al, and the Katz nomogram.

In this analysis, Gabrielle Werkoff, MD, from the department of obstetrics and gynecology, Hospital Tenon, Paris, and colleagues conducted a prospective multicenter study that compared the three models using data from 536 patients with breast cancer who had one to three metastatic sentinel nodes.

Of the 536 patients, 10.6% had more than four involved lymph nodes. The area under the curve for the Katz nomogram was 0.84 compared with 0.75 for the Louisville score excluding method of detection, and 0.77 for the Louisville score including the method of detection.

Results of the study were compared against an optimal logistic regression model. The optimal model scored 0.86 on the AUC compared with the Katz nomogram of 0.84.

The false negative rate for the Louisville score excluding method of detection was 1.8%; for Louisville including method of detection it was 9%; for the Katz nomogram it was 2.5%.

“In this study, the three models gave a false negative rate of less than 5%. This confirms their excellent performances,” wrote Lisa K. Jacobs, MD, and Charles M. Balch, MD, of the departments of surgery and oncology at Johns Hopkins Medical Institutions, and Seng-jaw Soong, MD, of the Comprehensive Cancer Center, University of Alabama at Birmingham. They wrote that this analysis by Werkoff and colleagues validates the clinical use of these three models in practice, but added that “although the Katz nomogram appeared to perform better than the two other models, it does not warrant routine usage in clinical practice” due to its moderate sensitivity and specificity.

“While the Katz nomogram appears to provide high negative predictive probability, the positive predictive value of only 23% is disconcerting,” they wrote. “This is especially true if adherence to this nomogram results in denying the use of immediate reconstruction in a large number of patients who would actually have been eligible for it.”

Therefore, although the use of the nomogram and other predictive models will grow in importance, for now “the use of intraoperative assessment and decision-making based on the existing nomograms appears for now to be limited,” they wrote.

Werkoff G. J Clin Oncol. 2009; doi:10.1200/JCO.2009.21.9139.