Abstract
Introduction
Multiple Myeloma (MM) is a heterogeneous disease with a strong need for robust markers for prognosis. Frequently occurring chromosomal abnormalities, such as t(4;14), gain(1q), and del(17p) etc. have some prognostic power, but lack robustness across different cohorts. Alternatively, gene expression profiling (GEP) studies have developed specific high risk signatures such as the SKY92 (EMC92, Kuiper et al. Leukemia 2012), which has shown to be a robust prognostic factor across five different clinical datasets. Moreover, studies comparing prognostic markers have indicated that the SKY92 signature outperforms all other markers for identifying high risk patients, both in single and multivariate analyses. Similarly, when assessing the prognostic value of combinations of various prognostic markers, the SKY92 combined with ISS was the top performer, and also enables detection of a low risk group (Kuiper et al. ASH 2014). Here, we present a further validation of the low and high risk groups identified by the SKY92 signature in combination with ISS on two additional cohorts of patients with diverse treatment backgrounds, containing newly diagnosed, previously treated, and relapsed/refractory MM patients.
Materials and Methods
The SKY92 signature was applied to two independent datasets. Firstly, the dataset from the Total Therapy 6 (TT6) trial, which is a phase 2 trial for symptomatic MM patients who have received 1 or more prior lines of treatment. The TT6 treatment regime consists of VTD-PACE induction, double transplant with Melphalan + VRD-PACE, followed by alternating VRD/VMD maintenance. Affymetrix HG-U133 Plus 2.0 chips were performed at baseline for n=55 patients, and OS was made available previously (Gene Expression Omnibus identifier: GSE57317). However, ISS was not available for this dataset. Secondly, a dataset of patients enrolled at two hospitals in the Czech Republic, and one in Slovakia (Kryukov et al. Leuk&Lymph 2013). Patients of all ages, and from first line up to seventh line of treatment were included (treatments incl Bort, Len, Dex). For n=73 patients Affymetrix Human Gene ST 1.0 array, OS (n=66), and ISS (n=58) was made available previously (ArrayExpress accession number: E-MTAB-1038). Both datasets were processed from .CEL files by MAS5 (TT6), and RMA (Czech), followed by mean variance normalization per probeset across the patients. The SKY92 was applied as previously described (Kuiper et al. Leukemia 2012), and identifies a High Risk and Standard Risk group. In conjunction with ISS, the SKY92 Standard Risk group is then further stratified into low and intermediate risk groups (Kuiper et al. ASH 2014). Kaplan-Meier plots were created, and the Cox proportional hazards model was used to calculate Hazard Ratios (HR), and associated 1-sided p-values that assess whether the SKY92 High Risk group has worse survival than SKY92 Standard Risk group (i.e. HR>1).
Results
Figure 1 shows the Kaplan Meier plots of the SKY92 High Risk and Standard Risk groups on the TT6 and Czech cohorts. On the TT6 dataset, the SKY92 signature identifies 11 out of 55 patients (20%) as High Risk. In both datasets, the SKY92 High Risk group has significantly worse overall survival, HR=10.3, p=7.4 * 10-6 (TT6), and HR=2.6, p=2.2 * 10-2 (Czech). In addition, the combination of SKY92 with ISS on the Czech dataset identifies a low risk group of 14 out of 61 patients (23%), with a five year overall survival estimate of 100% versus 28.7% in the SKY92 High Risk group (HR=inf). Robustness of the SKY92 signature is further demonstrated by the fact that it validates on both datasets, despite different microarray platforms being used.
Conclusions
The SKY92 high risk signature has been successfully validated on two independent datasets generated using different microarray platforms. In addition, on the Czech data, the low risk group (SKY92 Standard Risk combined with ISS 1) has been successfully validated. Together, this signifies the robust nature of the SKY92 signature for high and low risk prediction, across treatments, and with applicability in newly diagnosed, treated, and relapsed/refractory MM patients.
Figure 1. Kaplan-Meier plots showing a significantly poorer overall survival in patients identified as SKY92 High Risk (red curves), relative to SKY92 Standard Risk, on both the TT6 (left), and Czech (middle) datasets, as well as a low risk group by SKY92 & ISS1 on the Czech dataset (green curve, right). Figure 1. Kaplan-Meier plots showing a significantly poorer overall survival in patients identified as SKY92 High Risk (red curves), relative to SKY92 Standard Risk, on both the TT6 (left), and Czech (middle) datasets, as well as a low risk group by SKY92 & ISS1 on the Czech dataset (green curve, right).
Disclosures
van Vliet: SkylineDx: Employment. Ubels:SkylineDx: Employment. de Best:SkylineDx: Employment. van Beers:SkylineDx: Employment. Sonneveld:Celgene: Honoraria, Research Funding; Amgen: Honoraria, Research Funding; Karyopharm: Research Funding; SkylineDx: Membership on an entity's Board of Directors or advisory committees; Janssen: Honoraria, Research Funding.
Identification of Candidate Genes in Early-Stage Invasive Ductal Carcinoma Patients with High-Risk Mortality Using Genes Commonly Involved in Breast Cancer: A Retrospective Study