Premium
Prognostic evaluation in multiple myeloma: an analysis of the impact of new prognostic factors
Author(s) -
Ingemar Turesson,
Niels Abildgaard,
Tomas Ahlgren,
Inger Marie S. Dahl,
Erik Holmberg,
Martin Hjorth,
Johan Lanng Nielsen,
Anders Odén,
Carina Seidel,
Anders Waage,
Jan Westin,
Finn Wisløff
Publication year - 1999
Publication title -
british journal of haematology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.907
H-Index - 186
eISSN - 1365-2141
pISSN - 0007-1048
DOI - 10.1046/j.1365-2141.1999.01651.x
Subject(s) - medicine , percentile , beta 2 microglobulin , proportional hazards model , multivariate analysis , population , oncology , multivariate statistics , prognostic variable , multiple myeloma , survival analysis , gastroenterology , statistics , mathematics , environmental health
We have analysed the prognostic information for survival of presenting features in an unselected series of 394 myeloma patients. 15 variables with significant prognostic information were identified, among these were some not previously or only recently reported: serum levels of hepatocyte growth factor (HGF), interleukin‐6 (IL‐6), C‐terminal cross‐linked telopeptide of collagen I (ICTP) and soluble interleukin‐6 receptor (sIL‐6R). In a multivariate Cox analysis six variables were significantly and independently associated with poor survival: high age, low W.H.O.‐performance status (PS), high serum levels of calcium, β‐2‐microglobulin (β‐2M), IL‐6 and sIL‐6R. A risk score formed to predict survival for each percentile of the patient population allowed an efficient separation of prognostic groups. The discriminating power of the model compared favourably with three other previously published staging systems applied to the study population. Exclusion of IL‐6 and sIL‐6R from the model only marginally decreased the efficacy of the separation. The predictive value of some variables (sIL‐6R, β‐2M and W.H.O.‐PS) decreased significantly over time. We conclude that formation of a risk score based on independent variables is an efficient way to separate prognostic groups, that the contribution of new and not easily available parameters should be thoroughly evaluated before inclusion in prognostic models for clinical use and that the predictive value of parameters may decrease over time.