Reply:

Quinn, in his comments on the pathologic diagnostic procedures described in our paper, distorts our position. He ignores part of our record of published criteria and methods, and criticizes our discussion of the extant literature in regard to synaptophysin immunostains of normal and abnormal neuronal cells in tumors of the spinal cord. An unfortunate set of typographic errors in our citations, missed in editing the proofs, add to this problem, although the text should have made our intentions clear. We appreciate the opportunity to correct these citation errors and clarify the record.

Quinn states that we “fail to appropriately cite” the Zhang and Rosenblum study of synaptophysin staining in normal human spinal cord (1). In fact, in the Discussion, we clearly indicate on page 884, the second column, “Recently, the diagnostic utility of this pattern [of synaptophysin immunostaining] has been questioned by Rosenblum and others.” Our references 51 and 53, however, have unfortunate errors. We apologize for these, and regret any confusion they may have created. Reference 51 is to a chapter by Rosenblum in a major surgical pathology text. In the reference list, the citation is correct except that the year is given as “1966” when it should be “1996.” Reference 53 should have referred to the Zhang and Rosenblum article cited by Quinn (1). That this was our intention was clearly understood by Dr. Norenberg in his accompanying commentary (2) in which he also discusses this controversy, and notes, “while Patel et al reasonably address these concerns, appropriate caution is still necessary.”

Otherwise, we stand by our diagnoses. As noted, the tumors diagnosed come from a large series of pediatric intramedullary spinal cord tumors. The data from our histologic analyses have been presented at a meeting (3), but are currently still under review, and thus remain unpublished. Two neuropathologists (Douglas C. Miller, Lucy B. Rorke) evaluated all of the cases, and agreed on the diagnoses. Most of the diagnoses of ganglioglioma were made prior to examination of synaptophysin stains, and these stains were adjunctive or confirmatory in this majority of the cases we report. We found, as stated in our paper, that the radical resections of these tumors by our senior neurosurgical author (Fred Epstein) provided us with much more tissue from spinal cord tumors than is typically made available for pathologic analysis at other centers. This was of great importance in recognizing gangliogliomas. Quinn expresses doubt that neoplastic neuronal cell clustering in gangliogliomas occurs sufficiently often to obscure the diagnosis in smaller samples and cites one study (4) to affirm his position. As cited in our paper, most published series of gangliogliomas point to the tendency for the neuronal cells to cluster within entire zones lacking significant neuronal elements. Our references include the relatively large sets published by Diepholder and Isimbaldi (5, 6), and the more recent ones by Hirose and Büttner (7, 8). The Wolf article Quinn cites (4) in fact states that in 21 (33%) of 64 tumors the neurons were sparse, recognizing neuronal cells by special staining aided the diagnosis of ganglioglioma. These included immunostains for synaptophysin.

We emphasized the importance of radical resection specimens to the diagnosis of gangliogliomas in our original series (9); 20 biopsies were performed outside of our center. Biopsies showed only pure gliomas, and the identification of the tumors as gangliogliomas followed examination of radical resection specimens obtained at New York University Medical Center. In our discussion we explicitly stated that 71% of radically resected gangliogliomas in our series were identified by hematoxylin-eosin staining alone, whereas even with our own pathologic review, only 9% of biopsies ultimately shown to come from gangliogliomas were correctly identified from the biopsy itself. We concluded that, “the increased sensitivity and accuracy of histopathologic diagnosis provided by the synatophysin immunostaining pattern, and extensive sampling were both important in diagnosing gangliogliomas.”

Furthermore, we carefully stated that the perikaryal surface immunoreactivity for synaptophysin we used to diagnose gangliogliomas was “virtually unique” for gangliogliomas, and we noted the limited exceptions in the spinal cord by citing Zhang and Rosenblum (1). We noted identical results that have been reported by others (5, 6, 10), and can now add a study by Wolf et al (4) and two newer studies, by Hirose et al and Büttner et al (7, 8). Quinn misconstrues the term “virtually unique” for “literally unique” in his current letter.

Quinn's own studies, as reflected in the illustrations accompanying his letter, do not distinguish cytoplasmic immunoreactivity from this perikaryal surface pattern (see his Fig. 1A, C, and D). We have never used cytoplasmic staining for the diagnosis of ganglioglioma because it has long been clear that any interruption in axonal transport in native neurons results in perikaryal and axonal cytoplasmic immunopositivity for synaptophysin. For Quinn to use these “positive” neurons as part of his argument is inappropriate. Our observations on a much larger number of specimens do not match Quinn's findings of neuropil positivity around “normal” neuronal nuclear clusters in the cord or stem. He has furnished only high-power photomicrographs; there is no way to verify his observations from his figures, except to note that Figure 1A has an immunopositive neuropil.

One of the main points we have consistently made about synaptophysin immunostaining for the diagnosis of ganglioglioma is that, in some cases, the large cells in the tumor samples are originally thought to be large tumor astrocytes, and it is only with the immunostains that they are recognized as neurons. The issue of trapped normal neurons in a pure glioma vs. neoplastic neurons of a ganglioglioma can indeed be difficult, but in our experience, was rarely a problem. For Quinn to express such doubt about our entire set of diagnoses based on an argument that synaptophysin-positive neurons might be normal trapped elements is a gross misperception of the nature of the pathologic materials on which our diagnoses are based.

Quinn expresses surprise at the number of gangliogliomas we reported, which “triple the published cases extant only a few years ago.” As Quinn well knows, the senior neurosurgical author (Fred Epstein) has a unique practice with an international referral base for which he performs a large number of spinal cord intramedullary tumor resections each year. Indeed, from 1978 through 1994 there were 226 such operations on 174 children alone, with a substantial additional number not yet quantified from our records in adults. We have viewed our opportunity to review these pathologic specimens as a unique opportunity, as this must be the largest such collection of surgically resected intramedullary cord tumors anywhere.