Response to the letter to the editor: 1q gain is a frequent finding in preoperatively treated Wilms tumors, but of limited prognostic value for risk satisfaction in the SIOP2009/Gesellschaft für Pädiatrische Onkologie und Hämatologie (GPOH) trial

We were interested to read the letter by Vokuhl et al. (2014), submitted as a response to our research article, “Gain of 1q is a Marker of Poor Prognosis in Wilms Tumours” (Segers et al., 2013). We would like to take this opportunity to elaborate on the differences between the two publications, which are likely to explain their different conclusions, and to comment briefly on the limitations of post hoc prognostic biomarker studies performed on small series of samples. One striking feature of the Vokuhl et al. report is the proportion of samples found to have 1q gain, 45% in the test set and 58% in the validation series. These figures are substantially higher than has been reported in previous studies of significant numbers of tumors analyzed for genomic copy number variation by a variety of techniques, including the 19% found by conventional karyotyping in our own study (Segers et al., 2013) and the 27% found by multiplex ligation-dependent probe amplification (MLPA) by the Children’s Oncology Group (Gratias et al., 2013). Only where a sample series has been selected specifically to represent poor outcome cases have similar high levels of 1q gain been reported (Hing et al., 2001; Natrajan et al., 2006). The test and validation sample series analyzed by Vokuhl et al. were both significantly enriched in samples from high risk histological subtypes [50% and 26% respectively, compared to 12% of all Wilms tumors registered in the International Society of Paediatric Oncology (SIOP) Wilms tumour (WT) 2001 trial]. However, this does not fully explain the reported levels of 1q gain, which were even higher in the validation series (59%) than in the test series (45%). The authors suggest that this elevated frequency of 1q gain may be a result of analyzing only samples treated with preoperative chemotherapy (which might tend to select for subclones with gain) or may be due to their use of macrodissection for viable tumor. However, additional technical differences need to be considered before reaching any such conclusion. The report by Vokuhl et al. does not give any methodological details of their MLPA assay, in particular how they defined the threshold for “calling” 1q gain, nor any comparisons with alternative methods of copy number determination on the same samples. We have experience of applying the MLPA technique to both frozen and formalin-fixed Wilms tumors on which we have also assayed copy number variation using microarray single nucleotide polymorphism (SNP)-based platforms. We have recently participated in an international multicentre MLPA analysis of nearly 700 Wilms tumor samples treated with preoperative chemotherapy under SIOP protocols, described in a recent conference report (Chagtai et al., 2014); a manuscript describing the complete series is in preparation by members of the SIOP Renal Tumour Study Group. All samples were derived from frozen tissue and, although we did not macrodissect the specimens, we were guided by their histopathology, rejecting any that were known to be significantly contaminated with nontumor cells. We are therefore confident that our results are representative of the predominant 1q copy number status of the tumor samples we examined. In this MLPA-based analysis, 28% of the tumors had 1q gain. This is somewhat higher than the figure of 19% detected by cytogenetic analysis in our previous report (Segers et al., 2013) where we noted that our ability to detect all copy number events may have been reduced by fibroblast overgrowth in the cell cultures prior to karyotyping. It is, however, in line with the results of a recent large-scale MLPA study of an immediate nephrectomy series (Gratias et al., 2013), and is less than half the figure reported by Vokuhl et al. in their validation series. While it is possible their series may genuinely have contained an unusually high frequency of tumors with 1q gain, we believe this is unlikely given the similar genetic background of the majority population in Germany and the UK. In our