| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
EDITORIALS |
From the Department of Surgical Oncology, Princess Margaret Hospital, Toronto, Ontario, Canada.
Correspondence: Address correspondence to: David R. McCready, MD, Dept. of Surgical Oncology, Princess Margaret Hospital, 610 University Ave., Ste. 4-320, Toronto, Ontario, M5G-2M9, Canada.
In this issue of the Annals of Surgical Oncology, Zurrida et al.1 describe an exhaustive intraoperative method of analyzing sentinel lymph nodes (SLN) by using frozen section and immunohistochemistry for breast cancer patients. This methodology was developed because, in a previous cohort, frozen section of the sentinel nodes missed 32% of the metastases that were subsequently found on permanent analysis of these same nodes using multiple sections and immunohistochemistry.2,3 The authors wanted to reduce the subsequent need for axillary reoperation with its attendant psychological impact on the patient, and they are to be commended for their well recognized commitment to improving breast cancer care. However, before the clinical significance of this methodology can be judged, it should be analyzed for what it is–a new diagnostic test.
It would be easy to dismiss this frozen section methodology by claiming that the minimum 60 sections per node would be prohibitively expensive, that the extra 40 minute wait would be too costly in terms of surgical, operating room, and patient anesthetic time, and that the availability of an experienced, dedicated pathologist who could reliably process and read the slides during most breast cancer operations would be unlikely. Although these concerns essentially limit this technique to those with tremendous passion and research funds, there are other aspects of this diagnostic methodology that should be addressed.
The authors described three successive cohorts of patients who underwent SLN biopsy and simultaneous completion axillary dissection. The first 60 patients had no intraoperative assessment of the SLN and the false-negative rate of the SLN on permanent analysis compared with the remaining axillary nodes was 5% (2/37) and the prevalence of overall node positivity was 62% (37/60). The second group had a false-negative rate of 8% (7/88) and the prevalence of axillary nodal metastases was 46% (88/192). Intraoperative frozen section missed 32% of the SLN metastases. The last cohort consisted of the control arm within a randomized clinical trial and this group’s false-negative rate was 7% (11/154) and the prevalence of nodal disease was 41% (154/376). It should be noted that none of the above data was explicitly stated within the publication but can be derived from the given data. It was in this last group that the new exhaustive frozen section method was used.
The critical evaluation of a new diagnostic test should begin with the question: was there an independent, blinded comparison with a reference or "gold" standard?4 Unfortunately, this is not possible with the techniques described by Zurrida et al. because the frozen section evaluation used up all of the SLN material so that no tissue remained for paraffin blocks and permanent analysis. Thus, the evaluation of this new diagnostic method could not be compared with a reference standard. Would an independent observer, blinded to the frozen section results, obtain the same rate of nodal disease?
The authors acknowledged this deficit during their discussion but seemed to believe that because "the negative predictive value of the sentinel lymph node in the [third group] (95.3%) is higher than in the [second group] (93.7%) ..., this suggests that the new method is at least as reliable as permanent section analysis, if not more so." However, negative predictive values are derived values that are based on the prevalence of the quality/disease being tested as well as the inherent qualities of the diagnostic test (sensitivity, specificity, and, by definition, false-negative and false-positive rates). For example, if one assumes a constant false-negative rate of 7% (as in the article’s third group), the negative predictive value could range from 90% in a population with a 62% prevalence of node positivity (as in the article’s first group) to 98% if one examined a breast cancer population with only a 20% chance of nodal metastasis (e.g., a screened population). Therefore, for a given false-negative rate (1 - sensitivity) a lower prevalence of disease will result in a higher negative predictive value. It should be noted that the prevalence of node positivity declined in each successive cohort described in this paper and this includes a 23% (35/154) rate of micrometastatic disease in the last group. The point being that one cannot compare diagnostic tests using negative predictive values because the negative predictive value changes with prevalence.
The issue of exactly what is the reference standard in evaluating axillary dissections and sentinel nodes is important. When evaluating a diagnostic test, one should not arbitrarily change the gold standard, which for axillary nodes has traditionally been hematoxylin and eosin (H&E) sections from each lymph node block. The new test should be compared to an unwavering reference standard so as not to incur bias. One example of how the gold standard has shifted is in the permanent analysis of lymph nodes. When frozen section and cytologic touch preparations of lymph nodes are evaluated in relation to the traditional standard, the results are quite accurate. Fisher5 found that 90% of breast cancer nodal metastasis were identified by frozen section and Rubio6 reported a 96% rate using a touch preparation. If the reference standard is then altered to include multiple sections, and possibly immunohistochemistry, it seems quite logical that more disease will turn up in this intensively analyzed specimen as shown by Giuliano.7 If frozen sections are still performed with one or two slides from the center of the node, the sensitivity will be reduced when compared to the new reference of multiple sections. This should not be surprising and is exactly what the authors found in their second group. By empirically subjecting the sentinel node to multiple permanent sections and immunohistochemistry, the authors created a "decrease" in the accuracy of the traditional frozen section diagnosis. They then went on to repair the situation by analyzing the whole node by frozen section and eliminating the previously accepted gold standard.
This raises the question of the relevancy of multiple sections, immunohistochemistry, and micrometastatic disease in lymph nodes. I agree with a previous editorial8 that states that this process needs to be evaluated prospectively. Until the results of the National Surgical Adjuvant Breast and Bowel Project B-32 and American College of Surgeons Oncology Group Z-010 trials are available, we should not presume to know the prognostic meaning of micrometastatic disease. If one were to play the devil’s advocate, one could suggest that the micrometastatic disease found by Zurrida et al.’s intensive frozen section processing resulted in 35 patients who might have been over-treated as opposed to saving some patients a second procedure. Time, and the prospective clinical trials, will tell. For the time being, outside of a clinical trial, my belief is that touch prep or traditional frozen section of the sentinel node is reasonable and that permanent sections be performed using primarily H&E. Intensive sectioning and immunohistochemistry should only augment the H&E sections when appropriate.
Received for publication September 10, 2001. Accepted for publication September 20, 2001.
REFERENCES
This article has been cited by other articles:
 |
|
 |
|
  D. R. McCready Intraoperative Sentinel Node Assessment Is a Touchy Subject Ann. Surg. Oncol., May 1, 2002; 9(4): 321 - 323. [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
