Multiple Myeloma, Gammopathies

Gene signature combinations improve prognostic stratification of multiple myeloma patients

Abstract

Multiple myeloma (MM) is a plasma cell neoplasm with significant molecular heterogeneity. Gene expression profiling (GEP) has contributed significantly to our understanding of the underlying biology and has led to several prognostic gene signatures. However, the best way to apply these GEP signatures in clinical practice is unclear. In this study, we investigated the integration of proven prognostic signatures for improved patient risk stratification. Three publicly available MM GEP data sets that encompass newly diagnosed as well as relapsed patients were analyzed using standardized estimation of nine prognostic MM signature indices and simulations of signature index combinations. Cox regression analysis was used to assess the performance of simulated combination indices. Taking the average of multiple GEP signature indices was a simple but highly effective way of integrating multiple GEP signatures. Furthermore, although adding more signatures in general improved performance substantially, we identified a core signature combination, EMC92+HZDCD, as the top-performing prognostic signature combination across all data sets. In this study, we provided a rationale for gene signature integration and a practical strategy to choose an optimal risk score estimation in the presence of multiple prognostic signatures.

Introduction

Multiple myeloma (MM) is a B-cell neoplasm characterized by the accumulation of clonal plasma cells.1, 2 Studies over time have revealed both significant inter- and intra-clonal heterogeneity making personalized treatment selection challenging. Despite improvement in outcomes, myeloma is still not a curable disease.3, 4 With the underlying disease heterogeneity and increase in therapeutic options, it is useful to consider prognostic factors to identify patients with different disease course to better rationalize treatment.5

Over the years, a number of prognostic factors have been identified in myeloma. Studies have shown that gene expression signatures (GS’s) can effectively identify patients with very poor outcome, and numerous prognostic GS’s have been identified.6 Some of them were from direct comparison of patients with different survival,7, 8, 9 whereas others from underlying biology, which was subsequently found to be of prognostic importance.10, 11, 12, 13, 14, 15

Despite the potential utility, this approach has not been widely adopted in clinical practice as, although investigators have found powerful prognostic GS’s in myeloma, it is unclear how to make the most of them in clinical settings due to numerous unanswered questions such as their mutual relationship, the optimality of single signature versus combinations, and, in case of signature combinations, the possibility of outperforming combinations.

In this study, we tried to answer these questions through a series of detailed computational simulations of multiple GS combinations. By standardizing the estimation of GS index, we showed that combinations of signatures are better than any individual signatures in general. In the process, we developed a framework for incorporating new prognostic signatures into existing models for evaluation. Importantly, we found a specific signature combination that shows consistent improved performance across different data sets. This model can be easily incorporated into current analytic workflow and prospectively validated in clinical trials involving gene expression profiling (GEP).

Materials and methods

Data

We used three data sets from Gene Expression Omnibus (GEO):16 a GEP data set by the University of Arkansas Medical School researchers (UAMS; GSE2658),7 another from the APEX/SUMMIT/CREST clinical trial (APEX; GSE9782)17 and the third from the HOVON-65/GMMG-HD4 clinical trial (HOVON; GSE19784).9 For more details, please refer to Supplementary Methods. We downloaded data sets in SOFT format and used signal intensity in downstream analyses without additional processing.

GS and index estimation

We used the following GS’s that were significantly associated with MM survival in at least one of the data sets used in this study: chromosome instability signature (CINGECS),10 centrosome index (CNTI),11 prognostic signature by HOVON-65/GMMG-HD4 clinical trial researchers (EMC92),9 7-gene prognostic signature from MM cell line study (HMCL),12 cell death signature implicated by homozygous deletion (HZDCD),13 15-gene prognostic signature by Intergroupe Francophone du Myelome 99 clinical trial (IFM15),14 proliferation signature (PI),15 70- and 80-gene prognostic signatures by UAMS researchers (UAMS70, UAMS80).7, 8 We used the log2-transformed median-normalized expression values for index estimation. Please refer to Supplementary Methods for index estimation details.

GS integration, principal component analysis and survival analysis

In this study, we examined all non-redundant GS combinations; starting from each of nine single GS’s, we constructed all possible combinations of multiple signatures up to full nine signatures and performed survival analysis for each combination separately.

Three types of combination schemes were examined. In the first raw index integration scheme (scheme 1), indices of participating signatures were simply averaged without modification and used as indices of combined signatures. The second regularized index integration scheme (scheme 2) was identical with the first scheme except that each signature index was regularized to mean 0, s.d. 1 before averaging to balance the difference in index distribution. In the third dichotomized index integration scheme (scheme 3), each signature was segmented into two risk groups using signature-specific thresholds first (See Supplementary Methods), high and low index groups were given with 1 and 0 as risk scores for each signature, respectively, and component risk scores were added for each sample. This was equivalent to counting the number of high-risk signatures for a sample.

As a reference for comparison, we used the risk scores from the principal component analysis (PCA) of signature index matrix.18 The value along the first principal component, which determines the direction of maximum variance of aggregated gene signature indices was used as the risk score.

Survival analysis of either overall survival or progression-free survival as end points was performed using the Cox regression analysis. Hazard ratio (HR), 95% confidence interval and P-value, and the Harrell’s concordance index (C-statistic) of each combined GS were collected and analyzed.

Stratification of differential risk group based on GS index

Clinically, discrete risk group indicators are more convenient than continuous risk scores. To determine the optimal threshold for risk stratification, we carried out simulation studies for individual as well as combination signatures. For details, see Supplementary Methods and Supplementary Figures S4 and S7.

Code availability

Two R libraries, MMGEP and MMGeneSigIndex, that we developed ourselves were extensively utilized in this study.19, 20 For proper reproduction of results of this study, see Supplementary Data and Supplementary Methods.

Results

Patient characteristics were typical of myeloma and compatible

Patient characteristics of the three data sets considered were typical of myeloma (Supplementary Table S1). The APEX data set, comprising of late-stage patients, had higher plasma cell counts, and albumin, β2-microglobulin, C-reactive protein concentrations as expected. The composition of TC classes was also similar across data sets. UAMS and HOVON data sets displayed similar dispersed distribution of overall survival while APEX data set displayed rather distinct very narrow distribution of it (Supplementary Figure S1), reflecting the clinical difference between newly diagnosed and relapsed nature of respective patients.

Performance of different GS’s in different data sets

The distribution of each signature index was quite comparable among three data sets (Figure 1). These GS’s are moderately but significantly correlated (correlation coefficient: 0.3~0.6) except for HMCL7, which showed low correlation with other signatures (Supplementary Figure S2). Although most genes appear in only one of the signatures, there is significant overlap in high-risk patients of different signatures (Figure 2). Furthermore, the univariate survival analysis results indicated they were statistically significant prognostic markers despite changes in data sets and standardization of signature index estimation methods for unbiased index comparison, and hence suitable for constructing optimal prognostic models in myeloma (Supplementary Figure S3).

Figure 1
figure1

Distribution of gene signature indices across three data sets.

Figure 2
figure2

Overlap of signature-specific high-risk patients for UAMS data set under OS.

Combination of raw or regularized signature indices improved performance but not the dichotomized ones

Comparing the three signature integration schemes undertaken, we observed that increasing the number of member signatures improved performance in terms of HR (Figure 3) as well as survival test P-value and C-statistic (Supplementary Figure S5) in general when raw (scheme 1) or regularized signature indices (scheme 2) were combined. The performance improvement from GS integration, especially for HR, was more pronounced in scheme 1 than in scheme 2. However, when dichotomized signature indices were combined, increasing the number of component signatures was not beneficial (Figure 3c). Therefore, our subsequent analyses considered only integration schemes 1 and 2.

Figure 3
figure3

Performance of combined signatures with (top panel) raw (scheme 1), (middle panel) regularized (scheme 2) and (bottom panel) dichotomized gene signature index values (scheme 3). Horizontal lines correspond to the result of PCA-based survival analysis. Combinations of different number of signatures are marked with different colors for each combination of data set and survival type.

Simple averaging showed better performance than PCA-based risk scores

In combining GS’s, PCA is often used to extract the most important contributions from participating signatures.21, 22 In this study, we applied PCA to individual data sets separately and used the results as a reference to assess the performance of individual combinations (horizontal dashed lines in Figure 3) and full nine-signature combinations (Table 1). Although PCA-based risk scores were highly significant in various data sets, the unsophisticated average exhibited much higher HR and more significant P-value with comparable concordance. Also, almost all of GS combinations in schemes 1 and 2 showed performance better than PCA results.

Table 1 Comparison of survival analysis performance of two-signature combination EMC92+HZDCD (SigComb2) and nine-signature combinations (SigComb91, SigComb92)

Specific combination emerged from raw index integration

To examine if there were consistently outstanding combinations and how different integration schemes might affect the performance of these combinations, we examined the top-performing 10-signature combinations in terms of HR in detail (Supplementary Tables S2 and S3). Strikingly, the EMC92+HZDCD combination appeared repeatedly as parts of the list in raw index integration (scheme 1) though, upon signature index regularization (scheme 2), the benefit of unique high-performing combination waned and performance was gained through promiscuous combination.

Finally, we compared the performance of EMC92+HZDCD under integration scheme 1 (SigComb2) with full nine-signature combinations of schemes 1 (SigComb91) and 2 (SigComb92) to gain insight on the relative strength of carefully crafted two-signature combination (SigComb2) and simplistic ones (SigComb91, SigComb92). All risk scores were closely correlated with each other (Supplementary Figure S6) but SigComb2 showed much improved performance in terms of HR, P-value, and concordance in continuous scale (Table 1), and performed similarly to other risk scores in the separation of different survival groups under current scheme of differentiating risk groups (Figure 4 and Supplementary Figure S8). Although the signatures identified slightly varying fraction (9–23% depending on survival and data sets, respectively) of patients as risky (Supplementary Table S4), high-risk patients and the median survival time difference between low and high-risk groups for a fixed data set and survival type were similar (Supplementary Tables S4 and S5).

Figure 4
figure4

Survival curves of OS for EMC92+HZDCD (SigComb2) and full nine-signature combinations (SigComb91, SigComb92) under signature combination schemes 1 and 2 in UAMS, HOVON and APEX data sets. For UAMS and HOVON data sets, survival curves for UAMS70 (native signature of UAMS data set) and EMC92 (native signature of HOVON data set) are also shown, respectively, in orange for comparison.

Discussion

Research utilizing novel GEP technology has resulted in major new insights about the molecular biology of myeloma. However, as yet, these insights have not contributed to better outcomes or patient care. We believe that GEP will still prove useful in clinical settings, as it has been shown in many studies to provide powerful independent prognostic markers.

Several important points emerge from our studies that will have important implications on the clinical implementation of prognostic GS’s. First, if one has numerous prognostic signatures whose relative merits are uncertain (different validation populations, methodology and so on), then one can simply average all raw indices together to produce an effective risk score, as adding more signatures will generally improve performance compared with individual ones. This appears to be better than a sophisticated risk score based on a PCA-based projection method. Second, optimal and highly reproducible combinations persist across diverse data sets, suggesting that GEP can capture core biology that is not a result of random methodological artifact in different studies. As shown in our study, the EMC92+HZDCD combination provides highly improved performance compared with other signatures or combinations. Third, simple integration of dichotomized (such as low versus high-risk stratification) signatures is not beneficial. It is best to stratify patients into different prognostic groups after integrating raw indices.

Recently, researchers investigated the possibility of achieving improved prediction of patient survival by combining GS’s22, 23, 24, 25, 26 through 'meta-signature' approaches either by collecting core set of genes from those of baseline signatures26 or by sieving through comprehensive sets of GS’s or data.23, 24, 25 We took a different approach in this study based on a rationale that the publicly available and validated prognostic GS’s are likely to capture specific biological mechanisms responsible for MM progression. Given the multifaceted nature of MM biology, we considered it might be prudent to aggregate them together without discarding any of them to achieve maximum synergy. Furthermore, our method is quite intuitive and the analytical algorithm can be easily incorporated into GEP analysis workflow. As part of this effort, we have made the methodology and script for either combining EMC92+HZDCD or all signatures available online so that this can be easily utilized on occasions where GEP is available.19, 20

The evolution of prognostic factors in myeloma has already chronicled the incorporation of diverse developments from simple assays such as B2M, albumin, to their combination into the ISS, to the use of cytogenetics and fluorescence in situ hybridisation, and the subsequent combination of fluorescence in situ hybridisation with ISS and more recently the addition of LDH.27, 28, 29, 30 GS’s have very strong independent prognostic impact and our current study shows that combinations of these signatures (EMC92+HZDCD) are even stronger. The next step in the refinement of prognostication in myeloma will involve a comprehensive study looking at the contributions of ISS, genetics and GS combination to prognosis and the impact of their holistic combinations.31, 32

A number of caveats are worthy of consideration. The data sets we used included mainly patients of transplant eligible age. The fact that the EMC92+HZDCD combination is also able to discriminate survival in the APEX data set suggest that it may be broadly applicable, but the applicability of the scheme to elderly myeloma patients will require further study. Another is that we currently do not have clearly distinct therapeutic strategies for patients with different prognosis. This is an area of keen clinical interest especially for the very high-risk patients that are currently poorly served by even our best and newest treatments. At last, prognosis can be altered by different treatment and risk factors may differ depending on treatments given. In this regard, the availability of our combination schema for the analysis of GEP data from prospective trials using different drug combinations will allow easy assessment of whether new drugs and combination can alter the prognostic utility of GEP signatures.

In conclusion, we established that the combination of prognostic signatures is generally better than single signatures. The simple average of EMC92 and HZDC2 indices is the top-performing combination across data sets comprising of newly diagnosed and relapsed patients treated with novel agents and high-dose therapy with stem cell transplantation. This can be regarded as the standard prognostic GS with which future novel signatures need to be compared. The framework we introduced in this study can be adopted to assess future prognostic signatures. Hopefully, this sets the stage for the next step in the evolution of prognostic systems in MM, the incorporation of GEP to still an imperfect prognosis system despite recent addition of genetics to the ISS.

References

  1. 1

    Bergsagel PL, Kuehl WM . Molecular pathogenesis and a consequent classification of multiple myeloma. J Clin Oncol 2005; 23: 6333–6338.

    CAS  Article  Google Scholar 

  2. 2

    Laubach J, Richardson P, Anderson K . Multiple myeloma. Annu Rev Med 2011; 62: 249–264.

    CAS  Article  PubMed  Google Scholar 

  3. 3

    Pineda-Roman M, Zangari M, Haessler J, Anaissie E, Tricot G, van Rhee F et al. Sustained complete remissions in multiple myeloma linked to bortezomib in total therapy 3: comparison with total therapy 2. Br J Haematol 2008; 140: 625–634.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  4. 4

    San Miguel JF, Mateos M-V . Can multiple myeloma become a curable disease? Haematologica 2011; 96: 1246–1248.

    Article  PubMed  PubMed Central  Google Scholar 

  5. 5

    Fonseca R . Strategies for risk-adapted therapy in myeloma. Hematol Am Soc Hematol Educ Program 2007, 304–310.

    Article  Google Scholar 

  6. 6

    Stewart AK, Fonseca R . Prognostic and therapeutic significance of myeloma genetics and gene expression profiling. J Clin Oncol 2005; 23: 6339–6344.

    CAS  Article  Google Scholar 

  7. 7

    Shaughnessy JDJ, Zhan F, Burington BE, Huang Y, Colla S, Hanamura I et al. A validated gene expression model of high-risk multiple myeloma is defined by deregulated expression of genes mapping to chromosome 1. Blood 2006; 109: 2276–2284.

    Article  PubMed  Google Scholar 

  8. 8

    Shaughnessy JD, Qu P, Usmani S, Heuck CJ, Zhang Q, Zhou Y et al. Pharmacogenomics of bortezomib test-dosing identifies hyperexpression of proteasome genes, especially PSMD4, as novel high-risk feature in myeloma treated with Total Therapy 3. Blood 2011; 118: 3512–3524.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  9. 9

    Kuiper R, Broyl A, de Knegt Y, van Vliet MH, van Beers EH, van der Holt B et al. A gene expression signature for high-risk multiple myeloma. Leukemia 2012; 26: 2406–2413.

    CAS  Article  PubMed  Google Scholar 

  10. 10

    Chung T-H, Mulligan G, Fonseca R, Chng W-J . A novel measure of chromosome instability can account for prognostic difference in multiple myeloma. PLoS One 2013; 8: e66361.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  11. 11

    Chng W-J, Braggio E, Mulligan G, Bryant B, Remstein E, Valdez R et al. The centrosome index is a powerful prognostic marker in myeloma and identifies a cohort of patients that might benefit from aurora kinase inhibition. Blood 2008; 111: 1603–1609.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  12. 12

    Moreaux J, Klein B, Bataille R, Descamps G, Maiga S, Hose D et al. A high-risk signature for patients with multiple myeloma established from the molecular classification of human myeloma cell lines. Haematologica 2011; 96: 574–582.

    CAS  Article  PubMed  Google Scholar 

  13. 13

    Dickens NJ, Walker BA, Leone PE, Johnson DC, Brito JL, Zeisig A et al. Homozygous deletion mapping in myeloma samples identifies genes and an expression signature relevant to pathogenesis and outcome. Clin Cancer Res 2010; 16: 1856–1864.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  14. 14

    Decaux O, Lode L, Magrangeas F, Charbonnel C, Gouraud W, Jezequel P et al. Prediction of survival in multiple myeloma based on gene expression profiles reveals cell cycle and chromosomal instability signatures in high-risk patients and hyperdiploid signatures in low-risk patients: a study of the Intergroupe Francophone du Myelome. J Clin Oncol 2008; 26: 4798–4805.

    CAS  Article  PubMed  Google Scholar 

  15. 15

    Hose D, Reme T, Hielscher T, Moreaux J, Messner T, Seckinger A et al. Proliferation is a central independent prognostic factor and target for personalized and risk-adapted treatment in multiple myeloma. Haematologica 2011; 96: 87–95.

    Article  PubMed  Google Scholar 

  16. 16

    Barrett T, Keats JJ, Mittal V, Delmore JE, Zhang MQ, Moreau P et al. NCBI GEO: archive for functional genomics data sets—update. Nucleic Acids Res 2013; 41: D991–D995.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  17. 17

    Mulligan G, Mitsiades C, Bryant B, Zhan F, Chng W-J, Roels S et al. Gene expression profiling and correlation with outcome in clinical trials of the proteasome inhibitor bortezomib. Blood 2007; 109: 3177–3188.

    CAS  Article  Google Scholar 

  18. 18

    Bair E, Bourne PE, Weinmann AS, Chibon F, Smith AV, Nadon R et al. Prediction by supervised principal components. J Amer Statist Assoc 2006; 101: 119–137.

    CAS  Article  Google Scholar 

  19. 19

    R Library: MMGEP. Available from http://figshare.com/s/05c4814ca07d11e4872a06ec4bbcf141 (accessed 9 November 2015).

  20. 20

    R Library: MMGeneSigIndex. Available from http://figshare.com/s/580fd96aa07d11e4948106ec4b8d1f61 (accessed 9 November 2015).

  21. 21

    Bair E, Tibshirani R . Semi-supervised methods to predict patient survival from gene expression data. PLoS Biol 2004; 2: e108.

    Article  PubMed  PubMed Central  Google Scholar 

  22. 22

    Zhao X, Rødland EA, Sørlie T, Naume B, Langerød A, Frigessi A et al. Combining gene signatures improves prediction of breast cancer survival. PLoS One 2011; 6: e17845.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  23. 23

    Reyal F, Kerr MK, Maiso P, Tesarova L, Jiang J, Liu H et al. A comprehensive analysis of prognostic signatures reveals the high predictive capacity of the proliferation, immune response and RNA splicing modules in breast cancer. Breast Cancer Res 2008; 10: 1–15.

    Article  Google Scholar 

  24. 24

    Wirapati P, Sotiriou C, Kunkel S, Farmer P, Pradervand S, Haibe-Kains B et al. Meta-analysis of gene expression profiles in breast cancer: toward a unified understanding of breast cancer subtyping and prognosis signatures. Breast Cancer Res 2008; 10: R65–11.

    Article  PubMed  PubMed Central  Google Scholar 

  25. 25

    Buffa FM, Harris AL, West CM, Miller CJ . Large meta-analysis of multiple cancers reveals a common, compact and highly prognostic hypoxia metagene. Br J Cancer 2010; 102: 428–435.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  26. 26

    Abba MC, Lacunza E, Butti M, Aldaz CM . Breast cancer biomarker discovery in the functional genomic age: a systematic review of 42 gene expression signatures. Biomark Insights 2010; 5: 103–118.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  27. 27

    Durie BG, Salmon SE . A clinical staging system for multiple myeloma. Correlation of measured myeloma cell mass with presenting clinical features, response to treatment, and survival. Cancer 1975; 36: 842–854.

    CAS  Article  PubMed  Google Scholar 

  28. 28

    Greipp PR, San Miguel J, Durie BGM, Crowley JJ, Barlogie B, Bladé J et al. International staging system for multiple myeloma. J Clin Oncol 2005; 23: 3412–3420.

    Article  PubMed  PubMed Central  Google Scholar 

  29. 29

    Avet-Loiseau H, Attal M, Moreau P, Charbonnel C, Garban F, Hulin C et al. Genetic abnormalities and survival in multiple myeloma: the experience of the Intergroupe Francophone du Myélome. Blood 2007; 109: 3489–3495.

    CAS  Article  PubMed  Google Scholar 

  30. 30

    Avet-Loiseau H, Durie BGM, Cavo M, Attal M, Gutierrez N, Haessler J et al. Combining fluorescent in situ hybridization data with ISS staging improves risk assessment in myeloma: an International Myeloma Working Group collaborative project. Leukemia 2013; 27: 711–717.

    CAS  Article  PubMed  Google Scholar 

  31. 31

    Chng W-J, Dispenzieri A, Chim CS, Fonseca R, Goldschmidt H, Lentzsch S et al. IMWG consensus on risk stratification in multiple myeloma. Leukemia 2014; 28: 269–277.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

  32. 32

    Kuiper R, van Duin M, van Vliet MH, Broijl A, van der Holt B, Jarari el L et al. Prediction of high- and low-risk multiple myeloma based on gene expression and the International Staging System. Blood 2015; 126: 1996–2004.

    CAS  Article  PubMed  PubMed Central  Google Scholar 

Download references

Author information

Affiliations

Authors

Consortia

Corresponding author

Correspondence to W J Chng.

Ethics declarations

Competing interests

The authors declare no conflict of interest.

Additional information

International Myeloma Working Group

Niels Abildgaard1, Rafat Abonour2, Melissa Alsina3, Kenneth C Anderson4, Michel Attal5, Hervé Avet-Loiseau6, Ashraf Badros7, Nizar Jacques Bahlis8, Bart Barlogie9, Régis Bataille10, Meral Beksaç11, Andrew Belch12, Dina Ben-Yehuda13, Bill Bensinger14, P Leif Bergsagel15, Manisha Bhutani16, Jenny Bird17, Joan Bladé18, Annemiek Broijl19, Mario Boccadoro20, Jo Caers21, Michele Cavo22, Asher Chanan-Khan23, Ajai Chari24, Wen Ming Chen25, Marta Chesi26, J Anthony Child27, Chor Sang Chim28, Wee-Joo Chng29, Ray Comenzo30, Gordon Cook31, John Crowley32, Edvan Crusoe33, William Dalton34, Faith Davies35, Javier de la Rubia36, Cármino de Souza37, Michel Delforge38, Madhav Dhodapkar39, Meletios Dimopoulos40, Angela Dispenzieri41, Johannes Drach42, Matthew Drake43, Juan Du44, Brian GM Durie45, Hermann Einsele46, Theirry Facon47, Dorotea Fantl48, Jean-Paul Fermand49, Carlos Fernández de Larrea50, Rafael Fonseca51, Gösta Gahrton52, Ramón García-Sanz53, Laurent Garderet54, Christina Gasparetto55, Morie Gertz56, Irene Ghobrial57, John Gibson58, Peter Gimsing59, Sergio Giralt60, Hartmut Goldschmidt61, Jingli Gu62, Roman Hajek63, Izhar Hardan64, Parameswaran Hari65, Hiroyuki Hata66, Yutaka Hattori67, Tom Heffner68, Jens Hillengass69, Joy Ho70, Antje Hoering71, Jian Hou72, Jeffrey Huang73, Vania Hungria74, Shinsuke Ida75, Sundar Jagannath76, Andrzej J Jakubowiak77, Hans Johnsen78, Douglas Joshua79, Artur Jurczyszyn80, Martin Kaiser81, Efstathios Kastritis82, Jonathan Kaufman83, Michio Kawano84, Neha Korde85, Eva Kovacs86, Amrita Krishnan87, Sigurdur Kristinsson88, Nicolaus Kröger89, Shaji Kumar90, Robert A Kyle91, Chara Kyriacou92, Martha Lacy93, Juan José Lahuerta94, Ola Landgren95, Alessandra LaRocca96, Jacob Laubach97, Fernando Leal da Costa98, Jae-Hoon Lee99, Merav Leiba100, Xavier Leleu101, Suzanne Lentzsch102, Nelson Leung103, Henk Lokhorst104, Sagar Lonial105, Jin Lu106, Heinz Ludwig107, Anuj Mahindra108, Angelo Maiolino109, Elisabet E Manasanch110, Tomer Mark111, María-Victoria Mateos112, Amitabha Mazumder113, Philip McCarthy114, Jayesh Mehta115, Ulf-Henrik Mellqvist116, Giampaolo Merlini117, Joseph Mikhael118, Philippe Moreau119, Gareth Morgan120, Nikhil Munshi121, Hareth Nahi122, Weerasak Nawarawong123, Ruben Niesvizky124, Amara Nouel125, Yana Novis126, Michael O’Dwyer127, Peter O’Gorman128, Enrique Ocio129, Alberto Orfao130, Robert Orlowski131, Paula Rodriguez Otero132, Bruno Paiva133, Antonio Palumbo134, Santiago Pavlovsky135, Linda Pilarski136, Raymond Powles137, Guy Pratt138, Lugui Qui139, Noopur Raje140, S Vincent Rajkumar141, Donna Reece142, Anthony Reiman143, Paul G Richardson144, Joshua Richter145, Angelina Rodríguez Morales146, Kenneth R Romeril147, David Roodman148, Laura Rosiñol149, Adriana Rossi150, Murielle Roussel151, Stephen Russell152, Jesús San Miguel153, Rik Schots154, Sabina Sevcikova155, Orhan Sezer156, Jatin J Shah157, Kazuyuki Shimizu158, Chaim Shustik159, David Siegel160, Seema Singhal161, Pieter Sonneveld162, Andrew Spencer163, Edward Stadtmauer164, Keith Stewart165, Daryl Tan166, Evangelos Terpos167, Patrizia Tosi168, Guido Tricot169, Ingemar Turesson170, Saad Usmani171, Ben Van Camp172, Brian Van Ness173, Ivan Van Riet174, Isabelle Vande Broek175, Karin Vanderkerken176, Robert Vescio177, David Vesole178, Ravi Vij179, Peter Voorhees180, Anders Waage181, Michael Wang182, Donna Weber183, Brendan M Weiss184, Jan Westin185, Keith Wheatley186, Elena Zamagni187, Jeffrey Zonder188, Sonja Zweegman189

1Syddansk Universitet, Odense, Denmark; 2Indiana University School of Medicine, Indianapolis, Indiana, USA; 3H. Lee Moffitt Cancer Center and Research Institute, Tampa, Florida, USA; 4Dana-Farber Cancer Institute, Boston, Massachusetts, USA; 5Purpan Hospital, Toulouse, France; 6University of Toulouse, Toulouse, France; 7University of Maryland, Baltimore, Maryland, USA; 8University of Calgary, Calgary, Canada; 9M.I.R.T. UAMS Little Rock, Arkanas, USA; 10Institute de Biologie, Nantes, France; 11Ankara University, Ankara, Turkey; 12University of Alberta, Alberta, Canada; 13Hadassah University Hospital, Hadassah, Israel; 14Fred Hutchinson Cancer Center, Seattle, Washington, USA; 15Mayo Clinic Scottsdale, Scottsdale, Arizona, USA; 16Carolinas Healthcare System, Charlotte, North Carolina, USA; 17Bristol Haematology and Oncology Centre, Bristol, United Kingdom; 18Hospital Clinica, Barcelona, Spain; 19Erasmus MC, Rotterdam, The Netherlands; 20University of Torino, Torino, Italy; 21Centre Hospitalier Universitaire de Liège, Liège, Belgium; 22Universita di Bologna, Bologna, Italy; 23Mayo Clinic, Jacksonville, Florida, USA; 24Mount Sinai Medical Center, New York, NY, USA; 25Beijing Chaoyang Hospital, Beijing, China; 26Mayo Clinic Scottsdale, Scottsdale, Arizona, USA; 27Leeds General Hospital, Leeds, United Kingdom; 28Department of Medicine, Queen Mary Hospital, Hong Kong; 29National University Health System, Singapore; 30Tufts Medical School, Boston, Massachusetts, USA; 31University of Leeds, United Kingdom; 32Cancer Research and Biostatistics, Seattle, Washington, USA; 33Faculdade de Ciências Médicas da Santa Casa de São Paulo, Brazil; 34H. Lee Moffitt, Tampa, Florida, USA; 35University of Arkansas for Medical Sciences, Little Rock, Arkansas, USA; 36Hospital Universitario La Fe, Valencia, Spain; 37Univeridade de Campinas, Caminas, Brazil; 38University Hospital Gasthuisberg, Leuven, Belgium; 39Yale Cancer Center, New Haven, CT, USA; 40University of Athens School of Medicine, Athens, Greece; 41Mayo Clinic, Rochester, Minnesota, USA; 42University of Vienna, Vienna, Austria; 43Mayo Clinic Rochester, Rochester, Minnesota, USA; 44Changzhen Hospital, Shanghai China; 45Cedars-Sinai Samuel Oschin Cancer Center, Los Angeles, California, USA; 46Universitätsklinik Würzburg, Würzburg, Germany; 47Centre Hospitalier Regional Universitaire de Lille, Lille, France; 48Socieded Argentinade Hematolgia, Buenos Aires, Argentina; 49Hopital Saint-Louis, Paris, France; 50Hospital Clínic de Barcelona, Barcelona, Spain; 51Mayo Clinic Scottsdale, Scottsdale, Arizona, USA; 52Karolinska Institute for Medicine, Huddinge, Sweden; 53University Hospital of Salamanca, Salamanca, Spain; 54Hopital Saint Antoine, Paris, France; 55Duke University Medical Center, Durham, North Carolina, USA; 56Mayo Clinic, Rochester, Minnesota, USA; 57Dana-Farber Cancer Institute, Boston, MA, USA; 58Royal Prince Alfred Hospital, Sydney, Australia; 59University of Copenhagen, Copenhagen, Denmark; 60Memorial Sloan-Kettering Cancer Center, New York, NY, USA; 61University Hospital Heidelberg, Heidelberg, Germany; 62The First Hospital, Sun Yat-Sen University, Guangdong, China; 63University Hospital Ostrava and School of Medicine OU, Ostrava, Czech Republic; 64Tel Aviv University, Tel Aviv, Israel; 65Medical College of Wisconsin, Milwaukee, Wisconsin, USA; 66Kumamoto University Hospital, Kumamoto, Japan; 67Keio University School of Medicine, Tokyo, Japan; 68Emory University, Atlanta, Georgia, USA; 69University of Heidelberg, Heidelberg, Germany; 70Royal Prince Alfred Hospital, Sydney, Australia; 71Cancer Research and Biostatistics, Seattle, WA, USA; 72Shanghai Chang Zheng Hospital, Shanghai, China; 73National Taiwan University Hospital, Taiwan; 74Clinica San Germano, Sao Paolo, Brazil; 75Nagoya City University Medical School, Nagoya, Japan; 76Mt. Sinai Cancer Institute, New York, New York, USA; 77University of Chicago, Chicago, Illinois, USA; 78Aalborg Hospital Science and Innovation Center, Aalborg, Denmark; 79Royal Prince Alfred Hospital, Sydney, Australia; 80University Hospital, Cracow, Poland; 81Royal Marsden Hospital, Sutton, Surrey, United Kingdom; 82University of Athens, Athens, Greece; 83Emory Clinic, Atlanta, Georgia, USA; 84Yamaguchi University, Ube, Japan; 85National Institutes of Health, Bethesda, Maryland, USA; 86Cancer Immunology Research-Life, Birsfelden, Switzerland; 87City of Hope, Duarte, California, USA; 88Karolinska University Hospital and Karolinska Institutet, Stockholm, Sweden; 89University Hospital Hamburg, Hamburg, Germany; 90Department of Hematology, Mayo Clinic, Minnesota, USA; 91Department of Laboratory Med. and Pathology, Mayo Clinic, Minnesota, USA; 92Northwick Park Hospital, London, United Kingdom; 93Mayo Clinic Rochester, Rochester, Minnesota, USA; 94Grupo Español di Mieloma, Hospital Universitario 12 de Octubre, Madrid, Spain; 95Memorial Sloan Kettering Cancer Center, New York, New York, USA; 96Divisione Universitaria di Ematologia, Torino, Itay; 97Dana-Farber Cancer Institute, Boston, Massachusetts, USA; 98Instituto Portugues De Oncologia, Lisbon, Portugal; 99Gachon University Gil Hospital, Incheon, Korea; 100Sheba Medical Center, Tel Hashomer, Israel; 101Hospital Huriez, CHRU Lille, France; 102Columbia University, New York, New York, USA; 103Mayo Clinic Rochester, Rochester, MN, USA; 104University Medical Center Utrecht, Utrecht, The Netherlands; 105Emory University Medical School, Atlanta, Georgia, USA; 106Peoples Hospital, Beijing University, Beijing, China; 107Wilhelminenspital Der Stat Wien, Vienna, Austria; 108University of California, San Francisco, San Francisco, California, USA; 109Rua fonte da Saudade, Rio de Janeiro, Brazil; 110MD Anderson Cancer Center, Houston, Texas, USA; 111Weill Cornell Medical College, New York, New York, USA; 112University Hospital of Salamanca-IBSAL, IBMCC (USAL-CSIC), Salamanca, Spain; 113NYU Comprehensive Cancer Center, New York, New York, USA; 114Roswell Park Cancer Center, Buffalo, New York, USA; 115Northwestern University, Chicago, Illinois, USA; 116Sahlgrenska University Hospital, Gothenburg, Sweden; 117University of Pavia, Pavia, Italy; 118Mayo Clinic Arizona, Scottsdale, Arizona, USA; 119University Hospital, Nantes, France; 120University of Arkansas for Medical Sciences, MyelomaInstitute for Research and Therapy, Little Rock, Arkansas, USA; 121Dana-Farber Cancer Institute, Boston, Massachusetts, USA; 122Karolinska University Hospital, Stockholm, Sweden; 123Chiang Mai University, Thailand; 124Weill Cornell Medical College, New York, New York, USA; 125Hospital Rutz y Paez, Bolivar, Venezuela; 126Hospital Sírio Libanês, Bela Vista, Brazil; 127National University of Ireland, Ireland; 128Mater University Hospital, Ireland; 129University Hospital of Salamanca-IBSAL, IBMCC (USAL-CSIC), Salamanca, Spain; 130University Hospital of Salamanca-IBSAL, IBMCC (USAL-CSIC), Salamanca, Spain; 131MD Anderson Cancer Center, Houston, Texas, USA; 132Clinica Universidad de Navarra, Navarra, Spain; 133Clinica Universitaria de Navarra, CIMA, Pamplona, Spain; 134University of Torino, Torino, Italy; 135Fundaleu, Buenos Aires, Argentina; 136University of Alberta, Alberta, Canada; 137Parkside Cancer Centre, London, England; 138Heart of England NHS Foundation Trust, England, United Kingdom; 139Institute of Hematology and Blood Diseases, Tianjin, China; 140Massachusetts General Hospital, Boston, Massachusetts, USA; 141Mayo Clinic, Rochester, Minnesota, USA; 142Princess Margaret Hospital, Toronto, Canada; 143Saint John Regional Hospital, Saint John, New Brunswick, Canada; 144Dana-Farber Cancer Institute, Boston, Massachusetts, USA; 145Hackensack University Medical Center, Hackensack, New Jersey, USA; 146Bonco Metro Politano de Sangre, Caracas, Venezuela; 147Wellington Hospital, Wellington, New Zealand; 148Indiana University, Indianapolis, Indiana, USA; 149Hospital Clinic, Barcelona, Spain; 150Weill Cornell Medical College, New York, New York, USA; 151University of Toulouse, Toulouse, France; 152Mayo Clinic, Rochester, Minnesota, USA; 153Clinica Universitaria de Navarra, CIMA, Pamplona, Spain; 154Universitair Ziekenhuis Brussel, Brussels, Belgium; 155Masaryk University, Brno, Czech Republic; 156Memorial Sisli Hastanesi, Istanbul, Turkey; 157MD Anderson Cancer Center, Houston, Texas, USA; 158Tokai Central Hospital, Kakamigahara, Japan; 159McGill University, Montreal, Canada; 160Hackensack University Medical Center, Hackensack, New Jersey, USA; 161Northwestern University, Chicago, Illinois, USA; 162Erasmus MC, Rotterdam, The Netherlands; 163The Alfred Hospital, Melbourne, Australia; 164University of Pennsylvania, Philadelphia, Pennsylvania, USA; 165Mayo Clinic Arizona, Scottsdale, Arizona, USA; 166Singapore General Hospital, Singapore; 167University of Athens School of Medicine, Athens, Greece; 168Italian Cooperative Group, Istituto di Ematologia Seragnoli, Bologna, Italy; 169University of Iowa Hospital and Clinics, Iowa City, Iowa, USA; 170SKANE University Hospital, Malmo, Sweden; 171Levine Cancer Institute/Carolinas Healthcare System, Charlotte, North Carolina, USA; 172Vrije Universiteit Brussels, Brussels, Belgium; 173University of Minnesota, Minneapolis, Minnesota, USA; 174Brussels Vrije University, Brussels, Belgium; 175Vrije Universiteit Brussels, Brussels, Belgium; 176Vrije University Brussels VUB, Brussels, Belgium; 177Cedars-Sinai Cancer Center, Los Angeles, California, USA; 178Hackensack University Medical Center, Hackensack, New Jersey, USA; 179Washington University School of Medicine, St. Louis, MO, USA; 180University of North Carolina, Chapel Hill, North Carolina, USA; 181University Hospital, Trondheim, Norway NSMG; 182MD Anderson Cancer Center, Houston, Texas, USA; 183MD Anderson Cancer Center, Houston, Texas, USA; 184Abramson Cancer Center, Philadelphia, Pennsylvania, USA; 185Sahlgrenska University Hospital, Gothenburg, Sweden; 186University of Birmingham, Birmingham, United Kingdom; 187University of Bologna, Bologna, Italy; 188Karmanos Cancer Institute, Detroit, Michigan, USA; 189VU University Medical Center, Amsterdam, The Netherlands

Supplementary Information accompanies this paper on the Leukemia website

Supplementary information

Rights and permissions

Reprints and Permissions

About this article

Verify currency and authenticity via CrossMark

Cite this article

Chng, W., Chung, TH., Kumar, S. et al. Gene signature combinations improve prognostic stratification of multiple myeloma patients. Leukemia 30, 1071–1078 (2016). https://doi.org/10.1038/leu.2015.341

Download citation

Further reading

Search

Quick links