Robert, K.

8599 Background: MM survival varies enormously from few months to more than a decade. Baseline prognostic factors (PF) have been identified that are linked to survival outcomes, especially B2M and albumin. MM survival plots are typically characterized by 3 segments of slopes with decreasing steepness, suggesting that different PF may govern these time segments. Methods: We therefore examined, in S9321, PF that were associated with overall survival (OS) from baseline (BL) and from 3 subsequent landmarks 3, 5, and 7 years later (LM3, LM5, LM7). With a median follow-up of 9 years, OS and EFS were virtually identical in transplant and standard therapy arms so that the data were pooled. Results: Median OS and EFS for 817 eligible patients are 48 and 22 months. The BL model identified age (>60yr), hemoglobin (<10g/gL), platelet count (<130,000/uL), albumin (<3.5g/dL), B2M (>3.5g/dL), calcium (>10mg/dL), creatinine (>2mg/dL), LDH (>190U/L), performance status (PS, >1), IL-6 level (>140mg/L) and plasma cell labeling index (PCLI, >1%) as being univariately associated with short OS, while age (p=0.002), platelet count (p=0.04), calcium (p<0.001), B2M (p=0.002), LDH (p=0.010) and PCLI (p<0.001) retained significance in multivariate analysis. For the LM3 analyses, age (p=0.006) and B2M (p=0.009) were independently associated with poor OS, whereas PCLI was the only PF associated with LM5 (p=0.026) and age with LM7 (p<0.001). While associated univariately with OS from BL, CRP, platelet count, hemoglobin, albumin, calcium, creatinine, LDH, IL6 and PS all failed to affect OS from later LM times. Conclusions: Whereas the initial 3 years of OS are governed by a combination of features related to disease aggressiveness (LDH, calcium, albumin, CRP), tumor burden (hemoglobin, platelet count) and host tolerance (PS, creatinine), B2M and PCLI as reflectors of tumor burden and proliferation have sustained long-term OS implications. The age factor figures in partly due to death from other causes. In an era of ever-improving survival outcomes, such LM-based PF analysis may reveal trial-specific differences and thus provide guidance toward individualizing therapy. [Table: see text]