Bone marrow involvement in children with NB occurs in about 50%. However, in patients with high-risk or metastatic disease BM involvement may be detected in up to 80% of patients (Tolbert and Matthay 2018). Using highly sensitive techniques, BM involvement at diagnosis was found in even more than 90% of stage 4 / M patients (Méhes et al. 2001). Earlier studies compared the use of BM aspirates with trephine biopsies and concluded that the highest yield of detection was reached with the combination of both procedures and that BM trephines are slightly more sensitive than aspirates (Favrot et al. 1986; Mills and Bird 1986; Aronica et al. 1998). The German trial NB 2004 recommended performing either four aspirates from four different sites or two aspirates plus two trephine biopsies in case the aspirates did not appear to be representative (Berthold 2004). In the past two decades, immunologically based methods for assessing BM involvement in NB have widened the spectrum of diagnostic tools (Ambros et al. 2003; Méhes et al. 2003; Beiske et al. 2005; Morandi et al. 2015; Druy et al. 2018; Popov et al. 2019).
In this retrospective, single-center study in one of the largest NB research and treatment institutions in Germany, we compared the results of three different methods for the detection of NB cells in BM in 81 patients. These included CM, FCM and AIPF. This combination of BM analysis has been the standard approach in our institution for more than ten years.
BM cellularity was reduced in 37.4% of the aspirates. This is likely due to the fact that all patients had undergone numerous, previous therapies before treatment in our institution. Nevertheless, a high number of cells were available for analysis. NB cells in BM were most often detected by AIPF (19.5%), followed by FCM (14.1%) and CM (8.7%). During the observation period, the number of patients with BM involvement decreased from 39.5–23.1%.
BM infiltration detected by FCM was in a range of 0.01–28.3% (median 0.045%) and by AIPF in a range of 0.000012-94.5% (median 0.0067%) and thus in most samples in a sub-microscopic level of detection. This is supported by the finding that FCM and AIPF were able to identify additional BM samples with NB in aspirates which were unremarkable in light microscopy. The limited accuracy and analytical sensitivity of CM have been described in previous years by several authors (Méhes et al. 2003; Burchill et al. 2017; Popov et al. 2019). However, FCM and AIPF did not show consistent results in each sample. Inconsistencies were particularly found when the degree of BM infiltration was extremely low in a range of 0.01% or less. As a single method, AIPF identified NB cells most frequently and the combination of AIPF and FCM had the highest yield of detection of NB cells in BM.
In 143 assessments with unremarkable results in MRI and mIBG, we identified only seven BM samples with NB cells (4.9%), and those in fact with FCM and AIPF. In patients, whose tumor treatment has been finished for more than two years, our current algorithm is therefore obtaining BM samples only if there is detectable disease in mIBG/MRI. When the imaging studies reveal unremarkable results, we desist from further invasive diagnostics and procedures involving radiation exposure.
The mortality in the study population was 56.8%. The presence of NB cells in BM identified by CM at study entry reflected a more advanced stage of disease with a poor prognosis. All of these patients did not survive. Mortality was also very high in patients with BM involvement detected only by FCM or AIPF (78.3%) which underlines the importance of these immunological methods.
Retrospective analyses carry several limitations. In recent years, molecular diagnostic tests have also been used to assess the infiltration of NB cells into BM. In 2009, the consensus criteria of the International Neuroblastoma Risk Group Task Force for the detection of NB cells in BM were published (Beiske et al. 2009). Immunocytology for GD2 and reverse transcriptase quantitative polymerase chain reaction (RTqPCR) for tyrosine hydroxylase were recommended. In 2017, the International Neuroblastoma Response Criteria Bone Marrow Working Group recommended the collection of both bilateral BM biopsies and aspirates and their analysis with immunohistochemistry for GD2 and RTqPCR for tyrosine hydroxylase and paired-like homeobox 2B (PHOX2B) (Burchill et al. 2017). Although there may be substantial agreement that biopsies should be added to analysis of BM aspirates, this statement relates to data obtained in the 1980s and 1990s when diagnostic tools to detect minimal disease in aspirates were not well established (Franklin and Pritchard 1983; Verdeguer et al. 1988; Ganick et al. 1988; Thiesse et al. 1991; Hedborg et al. 1992; Brodeur et al. 1993; Monclair et al. 2009). Bone marrow assessment with RTqPCR as recommended by these two guidelines is currently not common clinical practice in Germany. Therefore, a comparison between those approaches and the results of bone marrow assessment analyzed here is not yet possible.
An association of RTqPCR for three different parameters (tyrosine hydroxylase, PHOX2B, doublecortin) in BM at diagnosis with poor outcome was shown (Stutterheim et al. 2008; Viprey et al. 2014). However, the authors also stated that the results of trephine biopsies did not contribute additional information. Its sensitivity is limited and cytology and histology cannot demonstrate sub-microscopic involvement of the BM (Cheung et al. 1997; Stutterheim et al. 2008). Due to the limited number of trephine biopsies obtained in our patients, these were excluded from further analysis. However, we were able to demonstrate that FCM and AIPF detect BM involvement at a sub-microscopic level and that this has a clear and substantial impact on survival.
In summary, AIPF proved to be the single most sensitive method for the detection of BM involvement in NB. The presence of NB cells in BM only identified by AIPF or FACS was capable of identifying a high-risk population with a poor survival of only 17.4%.