The original study by S. Smith et al (Lancet Oncology, 2011) demonstrated that the group had developed a 20-gene gene expression model (GEM) with an area under the curve (AUC) of 0.67 (95% CI 0.60-0.75) for prediction of nodal disease at cystectomy in AUO-AB-05/95 (cell line). The cutoff system identified patients with high relative risk (1.74, 95% CI 1.03-2.93) and low relative risk (0.70, 95% CI 0.51-0.96) of node-positive disease using 3 different cohorts of clinically node-negative patients with MIBC undergoing cystectomy.
- Of note, the study is based on cystectomy specimens.
In this study, the authors aim to externally validate the 20-gene model in a 150-patient sample. All 150 patients underwent a diagnostic TURBT followed by RC and PLND. RNA was isolated and the expression level of the 20 genes was determined on a qRT-PCR platform - The qPCR data were set as unreliable if the Ct (count) value was outside a confidence interval of 95% within each transcript. Ultimately, only 139 samples were included – of which 91 were lymph node negative and 48 were LN+. Samples were excluded in some of the standards (housekeeping genes ACTB and HPRT) had two or more data points that were unreliable.
Essentially, at the end of the day, this was a negative study. The authors found that the 20-gene model could not replicate the findings of the Smith group. Univariate analysis of each individual gene did not identify any that correlated with presence of node-positive disease. The AUC of the risk score calculated by the 20-gene model was 0.54 (95% Confidence Interval: 0.44-0.65).
The authors subsequently briefly mention a separate analysis using the TCGA (The Cancer Genome Atlas), in which a similar negative result was found. In the TCGA data, none of the 20 genes was differentially expressed in node-negative versus node-positive patients. Mann Whitney-U-test, range p-values 0.07-0.93.
In discussion with the presenting author, this group was not specifically matched to the Smith study cohort. In fact, there was no effort to match it – which makes it more clinically relevant, as it is more generizable. When asked how she may account for the negative results, she mentioned the following:
- Original study was cystectomy specimens. This study looked at TURBT specimens.
- Different platform was utilized to analyze the specimens
- Their study cohort was not matched to the original study
Tumor heterogeneity may also account for the difference. As the TCGA data was also negative, we can feel more confident that this gene model is not likely to make it to clinical practice.
It is refreshing to see the occasional negative study. With a lot of biomarkers being developed, validation has become much more important.
Smith SC, Baras AS, Dancik G, Ru Y, Ding KF, Moskaluk CA, Fradet Y, Lehmann J, Stöckle M, Hartmann A, Lee JK, Theodorescu D. “A 20-gene model for molecular nodal staging of bladder cancer: development and prospective assessment.” Lancet Oncol. 2011 Feb;12(2):137-43. doi: 10.1016/S1470-2045(10)70296-5. Epub 2011 Jan 20.
Presented by: Kim Van Kessel
Co-authors: Van De Werken H., Lurkin I., Ziel–Van Der Made A., Zwarthoff E., Boormans J.
1. Erasmus MC, Dept. of Pathology, Rotterdam, The Netherlands
2. Erasmus MC, Cancer Computational Biology Center (CCBC), Rotterdam, The Netherlands
3. Erasmus MC, Dept. of Urology, Rotterdam, The Netherlands
Written by: Thenappan Chandrasekar , Clinical Fellow, University of Toronto
at the #EAU17 -March 24-28, 2017- London, England