Clinicopathologic Factors Associated With False-Negative Sentinel Lymph-Node Biopsy in Breast Cancer

To the Editor:

The University of Louisville Sentinel Lymph Node Study (a multi-institutional prospective observational study) has made major contributions to our knowledge about the feasibility, indications, technique, and learning curve for sentinel lymph node (SLN) biopsy in breast cancer. Its particular strength is its breadth; by drawing on the experience of more than 300 surgeons performing more than 4000 SLN biopsy procedures in practice setting ranging from university teaching institutions to rural hospitals, it has been preeminent in establishing that SLN biopsy works in the “real world.”

SLN biopsy is falsely negative (FN) in about 5% of node-positive and (depending on the proportion of node-positive cases) in about 2% of all patients. A very large study is therefore required to identify with statistical significance those factors that might contribute to FN results. In a recent report,¹ the Louisville group have addressed this issue and in the process have raised as many questions as answers.

As they have shown before, FN were significantly more frequent when only 1 SLN was removed,² a strong argument for removal of all blue and hot SLN, not just the first one. As they have also shown before, FN were more frequent for inexperienced surgeons, and their present data interestingly suggest that the “learning curve” may constitute as few as 4 cases, compared with their earlier estimate of 20 cases.³ This result matches the experience of others⁴ and is consistent with a gradual maturation of the SLN biopsy technique.

Their observation that FN were more frequent for smaller tumors is harder to explain, and is discordant with their earlier finding that the FN rate was unrelated to tumor size.⁵ It is equally unclear why FN were more frequent with upper outer quadrant tumors. Finally, their observation that FN were more frequent with the use of immunohistochemical (IHC) staining makes no sense at all; increased pathologic scrutiny of the SLN can only act to increase the sensitivity of the examination (thereby decreasing the FN rate). This is confirmed by a detailed collective review of the SLN literature, comparing 27 studies using hematoxylin and eosin staining alone with 8 studies using hematoxylin and eosin plus IHC, with FN rates of 8% and 3%, respectively (P < 0.006).⁶ Since the Louisville study did not require a standardized pathology technique, it would be interesting to see a direct comparison of all clinicopathologic features for the IHC versus no-IHC cases to see if some other variable could explain this counterintuitive result.

The larger problem may lie elsewhere. Cases with incomplete data automatically “drop out” of multivariate analyses, and this may explain some of the perplexing results above. While their study purports to involve 4116 patients and 3869 successful SLN biopsy procedures by more than 300 surgeons, examination of the data tables indicates that while SLN and axillary node status were known in all cases, tumor size was missing in 33%, and all of the other clinicopathologic features were missing in 65% of cases. This pattern of missing data strongly suggests that the University of Louisville data consist of two distinct groups: one (comprising about 1340 patients) in whom complete and detailed information was collected for each case, and another (comprising the remaining two thirds) which recorded the barest minimum of data (SLN and axillary node status only). If, as appears to be the case, the authors are really reporting the results of a minority subset rather than the entire experience of their study, then how can we be sure that this subset is truly representative? Asked another way, are we seeing the collective results of more than 300 surgeons in a broad range of practice settings, or those from a selected subset of surgeons and institutions?

Hiram S. Cody, III, MD
Breast Service
Department of Surgery
Memorial Sloan-Kettering Cancer Center
New York, NY
codyh@mskcc.org