gms | German Medical Science

Neuroblastoma is the most common solid tumor in childhood, with an incidence of 1,3 in 100000 children per year. Metastasis to the bone marrow is a hallmark of high risk, predictive of poor outcome for most of the children. However, since the degree of infiltration is highly variable, it remains elusive whether single tumor cells are indicative of unfavorable prognosis either at diagnosis or during therapy. In order to improve the diagnosis of micrometastases and thus the risk prediction we compared different techniques on a series of 231 bone marrows from 136 patients: conventional cytology, immunocytology (anti-GD2) and RT-PCR (tyrosine hydroxylase). Samples were obtained from children at diagnosis (32%), during therapy (44%) and at relapse (24%) after up to 24h shipping. Patients were treated according to the national neuroblastoma trials. Cytologic evaluation revealed 33% of samples as positive (=infiltrated with tumor cells), whereas immunocytology (43%) and RT-PCR (45%) resulted in a higher percentage of positive specimens. Seventy-two percent of the positive samples showed an infiltration of below 1 tumor cell in 100 bone marrow cells, 13% contained up to 30% infiltration whereas in 15% hematopoesis was almost completely substituted by neuroblasts. Consistency of the results was 88% for the microscopy-based techniques, 81% for cytology and RT-PCR and 77% for immunocytology and RT-PCR. Discrepancies always regarded specimens with very low degree of infiltration (<1%) or even single tumor cells, which might in part be the result of sampling errors. For patients investigated at diagnosis the bone marrow result revealed no impact on the outcome, although the relatively short follow up time has to be considered. However, patients with still positive RT-PCR results during the first 4 months of therapy showed a worse outcome than patients with negative RT-PCR (N=18 versus N=16, P=0.009). Thus our results indicate that the use of immunocytology and PCR increases the number of positive samples and despite of the limited number and the short follow up time, it seems to improve the risk prediction.