- © 2000 by American Society of Clinical Oncology
Angiogenesis in Neuroblastoma
To the Editor:
We read with interest the recent study by Cañete et al1 evaluating angiogenesis in neuroblastoma. Unlike our initial study of 50 patients in which we found high tumor vascular index to be an independent adverse prognostic factor in neuroblastoma,2 these authors evaluated six different vascular parameters and did not find a correlation with clinical outcome. As appropriately stated by the authors, there were several differences between the two studies. However, we have several concerns regarding the study reported by Cañete et al.
Our first concern is that the population studied by Cañete et al is not a representative sample of neuroblastoma patients. The study popula- tion included 69 of 206 patients registered on the Spanish N-II-92 protocol over a 5-year period. The cohort selected for analysis largely consisted of patients with favorable biologic and clinical features. For example, although MYCN amplification is seen in approximately 20% of children with neuroblastoma, only three (5%) of the 58 patients in the population studied by Cañete et al had MYCN amplification. Furthermore, age, stage, and MYCN amplification were not significant prognostic indicators of survival in multivariate analysis, indicating that these patients do not constitute a representative sample of neuroblastoma patients.
We also have concerns regarding the amount of tumor tissue evaluated in the study of Cañete et al.1 The authors used the hot-spot technique to determine vascular parameters and analyzed an area of 0.59 mm2 per tumor; in our study, we evaluated the entire slide of each tumor, which averaged an area of 97.86 mm2. Although this allowed us to evaluate approximately 150 times more tumor specimen, the range of the number of vessels counted in the two studies was quite similar, suggesting a significant difference in the methodology employed in the two different studies. Weidner and Folkman3 have suggested that the “significance of the intratumoral microvessel density drops when the field size is smaller than 0.74 mm2.” In addition, the mean number of vessels counted by Cañete et al is greater than what has typically been seen in other tumors studied,2,3 suggesting potential underlying methodologic problems in their study.
Our third concern involves the use of the hot-spot technique in evaluating neuroblastoma tumor samples. A recent study by Tomlinson et al4 suggested that the “hot technique for counting vessels may not be appropriate in certain types of malignancies, such as sarcomas which are composed of a single malignant cell compartment.” In contrast, hot spot vessel numbers have been found to be predictive in the majority of carcinomas studied3 that are composed of two different cell compartments, one malignant and one stromal. Because the majority of neuroblastoma tumors are similar to sarcomas in that they are composed of sheets of tumor cells (ie, one malignant cell compartment), counting vessels in hot spots might not be reflective of the underlying tumor biology. In addition, evaluating only hot spots in neuroblastoma would enhance the likelihood of missing differences in vessel numbers that might exist in the more differentiated parts of a tumor.
Another potential source of error in the study by Cañete et al1 is the use of computerized image analysis. We are not aware that the fully computerized methods used to count vessels in this study have been completely validated. We are aware of the difficulties inherent in immunohistochemical stains, but the authors make little mention of this fact and how they controlled for it. Furthermore, they do not describe any of the samples being excluded because of staining difficulties. However, a recent consensus article5 on the quantification of angiogenesis questioned the role of computerized image analysis in part because the “high-signal to noise ratio of the immunostaining required… might result in a considerable number of tumor sections unsuitable for evaluation.”
Recent studies by Erdreich-Epstein et al6 have shown that integrins αvβ3 and αvβ5, markers of angiogenic endothelium, are more highly expressed in blood vessels in high-risk versus low-risk neuroblastoma tumors, further extending our studies. In addition, several preclinical studies have shown the efficacy of antiangiogenic therapy in treating neuroblastoma in animal models. Together, these studies strongly suggest that angiogenesis contributes to the clinical behavior of neuroblastoma tumors and that angiogenic inhibitors may be effective in the treatment of high-risk neuroblastoma patients.
References
Response
In Reply:
We are thankful for the opportunity to comment on the letter from Katzenstein et al regarding our study about vascular parameters in a cohort of neuroblastoma patients.1 When the study by Meitar et al2 was published in the Journal of Clinical Oncology in 1996, we were very interested in analyzing our patients with a more objective and reproducible method than the one proposed by these authors. At that time, the methodology to study vascular parameters in breast carcinomas by means of computerized methods was fully established and developed in our department of pathology. Indeed, that methodology followed the recommendations given by the consensus of experts,3 from which Katzenstein et al extracted only one sentence for their comment. Besides, this methodology has been published elsewhere.4
Cohorts of 50 or 66 patients in retrospective studies, as in these two, are not representative of neuroblastoma. In our case, the 66 patients did not statistically differ from the large cohort of Spanish patients. We have already communicated5 the favorable characteristics of Spanish neuroblastoma patients in the N-II-92 protocol. Although retrospective, it had the advantage that all patients were treated according to one protocol; therefore, differences in outcome would be due to tumoral behavior, abolishing the differences that could appear due to different therapeutic strategies over the years. From our study, we did not make conclusions about the prognostic impact of angiogenesis in neuroblastoma; however, in our group of patients, the prognostic impact of angiogenesis was not as crucial as we expected. We observed some differences in vascular parameters in relapses, but they were not statistically significant. We believe this important question can only be addressed in larger multicentric supranational studies, as it has been done for MYCN, 1p deletion, or other prognostic factors in neuroblastoma.
It is well known that neuroblastoma is a very complex and heterogeneous tumor from the pathologist’s point of view,6 with tumoral and stromal areas of very different appearance and behavior in the same slide. Vascular hot spots are the critical regions to assess accurately the association between tumor progression and the angiogenic potential; choosing them is the most critical point, and the training and experience of the investigator is very important. Pilot studies7 have suggested that the degree of subjectivity related to the hot spot selection step is acceptable. In our case, vascular hot spots were encountered predominantly at the peripheral tumor margin, at low magnification, with a low background staining and a highly specific and intense labeling of endothelial cells (anti-CD34, microwave antigen retrieval, and protease pretreatment). Indeed, three (4%) of 69 initial patients had to be excluded. That small percentage is due to the previous development and validation of the methodology at our laboratory.4 Although Weidner et al8,9 have suggested a field size of 0.74 mm2, a wide range has been used, from 0.12 to 1.74 mm2. The range of vessel number in our study varied between nine and 1,034, with the frequencies shown in Fig 1. Again, different studies2,8-10 have used methodologies that are different from ours and are not comparable.
Fig. 1. Hystogram of microvessel number in 66 samples analyzed.
Because of the pathologic complexity of neuroblastoma, it is not known whether the vasculature patterns in this tumor are similar to those of carcinomas or sarcomas. It could be that different vascular patterns in neuroblastoma exist, related perhaps to biologic aspects (more differentiated v undifferentiated) or location (abdominal v thoracic), as has been described for primitive neuroectodermal tumors,11 which are closer to neuroblastoma than sarcomas. A better and deeper knowledge of vasculature patterns in neuroblastoma is needed so that scientists can choose the right methodology for studying the prognostic impact of angiogenesis in neuroblastoma.
From our point of view, the main disadvantages of our methodology are as follows: first, it is time-consuming, and second, it requires sophisticated equipment that might not be available for all laboratories. In that sense, the methodology used by Meitar et al2 is more appealing and would be useful for our daily practice if it followed the consensus recommendations.3 As proposed before, any methodology used to assess angiogenesis in neuroblastoma should be validated in larger prospective supranational studies.
