Résumé
Background. Survival in AL-amyloidosis is thought to be primarily determined by signs and symptoms caused by deposition of amyloid light chains, most prominently in the heart. In contrast, molecular characteristics of the underlying malignant plasma cell disease have been described, but are mostly considered less important.
Aim of our study is to predict malignant plasma cell biology related survival in AL-amyloidosis by establishing the first gene expression based risk stratification (HDAL) and assessing its independence from clinical serum parameter assessing risk.
Methods. CD138+-purified plasma cell samples of 919 patients with malignant plasma cell diseases, i.e. 195 AL-amyloidosis and 724 symptomatic myeloma patients were investigated by gene expression profiling, 124 AL-amyloidosis patients by RNA-sequencing. Gene expression profiling data of AL-amyloidosis patients were spitted in a training (TG, n=99) and a validation group (VG, n=96). A two-step model according to Rème et al. was applied on the training group, including a running log rank test for gene selection and a multi-cut-off running log-rank algorithm for optimal cut-off-selection, leading to a selection of 15 genes associated with good and 44 genes with adverse prognosis. The resulting HDAL-score was validated on the independent VG and our myeloma patient cohort using survival estimates for censored data. Differences between curves were assessed using the Log-rank test. The continuous RNA-Seq HDAL-score (R-HDAL) was subsequently derived to validate the survival association of the selected genes. HDAL was tested for independence with serum parameter assessing clinical staging systems by multivariate Cox regression.
Results. Categorical split of the HDAL-score delineates three significantly predictive groups of 48%, 29% and 23% of AL-amyloidosis patients with a median survival of 105, 53 and 3 months in the training, and 72, 33 and 6 months in the validation group, respectively. In symptomatic myeloma patients, HDAL significantly stratifies two groups with a median survival of 128 and 78 months. Thus, HDAL is a malignant plasma biology related derived risk stratification. HDAL and R-HDAL are significantly predictive for survival as continuous parameters. Categorial and continuous HDAL are individually independent predictive from clinical staging systems, i.e. Mayo 2004, 2012 and Euro score, and the assessed serum parameters, i.e. NT-ProBNP, cTNT and dFLC.
In conclusion, malignant plasma cell biology related and amyloid deposition mediated survivals in AL-amyloidosis are independent. Prognosis driven by the first component can significantly be assessed by transcriptome profiling (HDAL or R-HDAL).
We thank U. Hegenbart and S. Schönland for clinical collaboration in this work.