Validation of Pre-Transplant Biomarker Measurement and Risk Score for Non-Relapse Mortality after Allogeneic Hematopoietic Cell Transplant (HCT)

A recent transplant registry study in patients with acute myeloid leukemia (AML) and myelodysplastic syndrome (MDS) found that pre-transplant serum c-reactive protein (CRP), ferritin, and albumin were associated with non-relapse mortality (NRM) and overall survival (OS) after unrelated donor HCT, and derived a predictive biomarker risk score (BRS) [ Haematologica 2016 101:1426]. We evaluated these biomarkers and BRS using an independent dataset and further characterized the clinical factors driving abnormal biomarkers.

Pre-conditioning levels of CRP, albumin, and ferritin were retrospectively reviewed in adults undergoing first allogeneic HCT (n = 359) at the University of Chicago between 2008 and 2014 for a variety of hematologic malignancies, conditioning regimens and donor sources (131 matched related, 153 matched unrelated, and 75 haplo-cord). The BRS was calculated using the published formula [figure]. We also examined the associations of individual biomarkers with clinical data measured in the 90-day (weight loss, red blood cell [RBC] transfusions, and bacteremia) or the 30-day pre-conditioning period (creatinine, liver aminotransferases, fever, comorbidity index score, disease risk, age, and performance status [PS]).

A variety of clinical factors showed significant but modest correlations with abnormal biomarkers. Factors associated with elevated CRP included higher comorbidity (r = 0.42), lower PS (r = 0.39), high risk disease (r = 0.29), and RBC transfusion (r = 0.25). Albumin correlated with high risk disease (r = 0.60), fever (r = 0.52), and prior bacteremia (r = 0.42). Ferritin correlated with elevated liver enzymes (r = 0.37), older age (r = 0.31), higher comorbidity (r = 0.31), and RBC transfusion (r = 0.27). In multivariate analyses, CRP > 10 mg/L (n= 92, 35.1%) predicted significantly increased NRM at 3 years (HR = 2.22, P <0.001), primarily mediated by early NRM (1 year HR = 2.54, P <0.001). The effects of albumin < 3.5 g/L (P = 0.23) and ferritin > 2500 ng/mL (P = 0.36) on 3-year NRM were not significant. Both CRP (HR = 1.63, P = 0.021) and albumin (HR = 1.92, P = 0.016) predicted inferior 3-year OS in multivariate analysis, while ferritin did not (P = 0.32). The high-risk BRS group (n = 31, 13.3%) had significantly worse 3-year OS (HR = 2.04, P = 0.023) and higher NRM (HR = 2.75, P = 0.002) compared to low and intermediate risk. However, low and intermediate BRS groups did not significantly differ in 3-year OS (P = 0.067) or NRM (P = 0.19). We also found no significant difference in overall NRM among all three groups [figure]. Similar outcomes were observed in the AML/MDS subset (n = 217) with a significant difference in the high-risk BRS group at 3 years (OS: HR = 2.44, P = 0.015; NRM: HR = 3.00, P = 0.004) but not between the intermediate and low risk groups (P = 0.13 for OS; P = 0.85 for NRM).

The modest associations of multiple clinical factors to serum biomarkers suggest the biomarkers encompass a non-specific but meaningful aggregate measure of pre-HCT vulnerabilities. In a heterogeneous population of diseases and donor sources, we confirmed pre-transplant CRP independently predicts for inferior OS through higher NRM. Only the high-risk BRS group (13% of cohort) experienced worse NRM and OS, whereas no difference in NRM was found among all three groups due to similar outcomes for the low- and intermediate-risk BRS groups. Pre-transplant albumin and ferritin did not predict NRM and therefore hampered the performance of the BRS. These data support incorporation of pre-transplant CRP, but not ferritin or albumin, as a biomarker for HCT risk-stratification. Future research should explore the mechanisms behind elevated CRP and elevated transplant associated mortality.

Figure: OS and NRM by pre-transplant CRP and BRS. (A) CRP >10 predicts a significant difference in OS by log-rank test in Kaplan-Meier analysis. (B) CRP >10 predicts a significant difference in NRM in competing risks subdistribution analysis. (C) BRS formula: 0.44633*I(CRP > 3.54) + [0.07184+0.19646*ln(Ferritin)] + 0.44730*I(albumin < 3.4), where "I" is an indicator for the parenthetical condition. Risk categories were defined as low (< = 1.5), intermediate (> 1.5 and < = 2.0), and high (> 2.0). There is a significant difference in OS among all BRS categories by log-rank test. (D) There is not a significant different in NRM among all BRS categories in competing risks analysis.

View largeDownload slide

View largeDownload slide

Close modal