Validation of SKY92 high and low risk prognostication in a retrospective, multinational cohort of 155 non-trial multiple myeloma patients
Yan-Ting Chen1, Erik T. Valent1, Erik H. van Beers1, Rowan Kuiper1, Stefania Oliva2, Torsten Haferlach3, Wee-Joo Chng4, Martin H. van Vliet1, Pieter Sonneveld5, Alessandra Larocca2
1SkylineDx, Rotterdam, The Netherlands
2Myeloma Unit, Division of Hematology, University of Turin, Turin, Italy
3MLL Münchner Leukämielabor GmbH, Munich, Germany
4National University Cancer Institute, National University Health System, Singapore, Singapore
5Department of Hematology, Erasmus MC Cancer Institute, Rotterdam, The Netherlands
Genetic hallmarks of multiple myeloma (MM) patients have been extensively studied based on biobanked samples that have been collected as part of clinical trials. For example, the gene expression profiling (GEP) based SKY92 signature, which identifies ~15-20% of patients as high risk, has been independently validated in more than ten trials as a prognostic factor for progression free survival (PFS), and overall survival (OS). Limited evidence is available outside of trials, and here we present the first analysis of SKY92 and other markers on real world MM data.
Methods: As part of the HORIZON2020 funded MMpredict project, a total of n=155 samples (136 newly diagnosed, 19 relapsed) from non-trial MM patients were collected from the National University Hospital System (Singapore, n=73), Münchner Leukämielabor (Germany, n=37) and the University of Turin (Italy, n=45). Median age was 65 years (range 32-90), and treatments were diverse. MMprofilers were performed on all samples, providing SKY92 risk status results. Interphase FISH, and ISS results were collected from health records. Hazard ratios (Cox proportional hazards model) were calculated and evaluated by the likelihood ratio test to assess the significance of each marker. We also stratified the risk groups by combining SKY92 and ISS as proposed before.
Results: There were 34/155 patients with SKY92 high risk (22%) with hazard ratio of 2.7 (p=6.5E-05 with 95% CI 1.66 – 4.41) for PFS and 3.2 (p=1.7E-04 with 95% CI 1.74 – 5.88) for OS. Other markers that were univariately significant for both PFS and OS, were: Clusters MF and PR, virtual gain(1q) and interphase del(13p). In a multivariate Cox proportional hazards model, SKY92 and Cluster PR remained significant for both PFS and OS among all univariately significant markers (interphase del(13p) excluded since 35% of data was missing). Surprisingly, ISS II compared to ISS I had very similar survival (PFS: HR 0.9, p=0.78; OS: 1.0, p=0.98), whereas ISS III had worse, but not significantly different, survival compared to ISS I (PFS: HR 1.2, p=0.56; OS: 2.2, p=0.15). The three class SKY92–ISS classifier identifies high risk (HR; SKY92 high risk), intermediate risk (IR; SKY92 standard risk and ISS II or III) and low risk (LR; SKY92 standard risk and ISS I). Using these classes for PFS resulted in IR vs LR (HR 1.77; p=0.28) and HR vs LR (HR 4.81; p=0.004). For OS the results were IR vs LR (HR 1.69; p=0.48) and HR vs LR (HR 6.47; p=0.001).
Conclusions: Real world evidence confirms that SKY92 is prognostic for PFS and OS in the non-trial setting. In addition, we found indications that despite the poor association of ISS with outcome, the LR class (SKY92 SR ISS I) tends toward the low risk in the non-trial setting as was seen previously.
Funding: This project has received funding from the European Union’s Horizon 2020 research and innovation programme under grant agreement No 701143