Science Spotlight

Patient factors predict outcomes for HCT

Kaplan-Meier survival plots for non relapse mortality (NRM) in lymphoma patients undergoing autologous Hematopoietic cell transfer (HCT) patients stratified by unique risk factors. Top, patients are dichotomized into yes or no groups for history of alcohol use disorder (AUD). History of AUD was the factor more strongly associated with NRM. Bottom, patients were stratified by number of risk factors: HCT-CI greater than three, age by decade, and history of AUD.

Hematopoietic cell transfer (HCT) is a lifesaving technology that since its inception has been optimized to treat many diseases. While HCT often yields cures, some complications still exist for this therapy including recurrence, graft versus host disease (GVHD), infection, and cardiopulmonary complications. Understanding risk factors for patients can help inform treatment selection and behaviors. Many studies have shown that the type of malignancy and the source of hematopoietic cells (self/autologous vs donor/allogenic) strongly influence patient outcomes; however, there is limited understanding of how patient-specific features/behaviors influence outcomes. Researchers in the Sorror Lab (Clinical Research Division) performed a retrospective study of lymphoma patients undergoing autologous HCT to understand which patient-specific factors contribute to non relapse mortality (NRM) and overall survival (OS). These findings were recently published in Biology of Blood and Marrow Transplantation.

Historically, patient risk has been evaluated using the HCT-specific comorbidity index (HCT-CI). This index is determined by weighting many features such as patient organ function, other cancers, medical comorbidities like infection, and limited psychological data. The functionality of this tool has not been well examined in lymphoma patients. In this study researchers validated the use of HCT-CI and searched for any other patient-specific factors that predict NRM outcomes.

The patient cohort analyzed for this study consisted of 754 individuals and was predominantly white (86%), married (69%), male (69%) never smokers (55%). The median age was 53 and HCT-CI ranged from 0-9 with a median of 1. Using this group researchers performed retrospective multivariate analysis to determine which factors correlated to NRM and OS. From these efforts researchers found three patient-specific factors that negatively correlated with NRM: HCT-CI, age (stratified by decade), and history of alcohol use disorder (AUD).

In the case of HCT-CI patients ranking three or greater on the index had higher NRM (P=0.05), yet scores 0-2 did not correlate. This finding was highlighted the need to verify the sensitivity of this index for unique patient types. A previous study using HCT-CI showed that any score greater than zero correlated to increased risk of NRM, yet that study was analyzing patients that underwent allogeneic HCT (from a donor) and this study only analyzed autologous HCT (from the patient). The relationship between age and NRM was also nuanced. Many studies have previously found that patient age does not affect treatment related mortality, and thus has limited predictive power for patient survival. In fact, if patients were separated into two groups based on the median age of 53 there was no association between age and NRM. Instead, researchers performed multivariate analysis on age stratified by decade and found that patient age per decade carried significant risk to NRM (P=0.02). Finally, researchers found the strongest association with NRM in this group was a history of AUD (P=0.004). One reason AUD was the strongest association in this cohort is because AUD is associated with male sex and depression features that were common to this study and HCT patients in general.

Using age, HCT-CI, and AUD status researchers assigned a 'risk factor score' to each patient ranging from zero to three risks – one for each association. When stratified by number of risk factors the association with NRM was much more prominent and had moderate predictive power as evidenced by the c-statistic estimate that reach 0.64 when all three risk factors were combined.

The most puzzling finding from this retrospective study was that the association between NRM and AUD was difficult to attribute to a specific pathology. A reasonable assumption would be patients with AUD would experience NRM from hepatic damage or immunosuppression. However, no strong correlation between liver enzyme function and NRM could be observed. It is possible subtle hepatic or immune damage was present but not observed in these patients. Possible evidence is that AUD was common in patients experiencing NRM from infection (26%) or pulmonary complication (38%).

As they stand, these risk factors do not inform risk-mitigation strategies, a study exploring such a relationship would be useful to patient counseling. As AUD was strongly associated with NRM such risk-mitigation studies could include treatment for depression and ensuring medicine compliance as both are come in patients with AUD.

Graf SA, Vaughn JE, Chauncey TR, Storer BE, Gopal AK, Holmberg LA, McCune JS, Bensinger WI, Maloney DG, Press OW, Storb R, Sorror ML. 2016. Comorbidities, Alcohol Use Disorder, and Age Predict Outcomes Following Autologous Hematopoietic Cell Transplantation for Lymphoma. Biol Blood Marrow Transplant. Epub ahead of print.

Funding for this research was provided by the National Institutes of Health, the American Cancer Society, and the Patient-Centered Outcomes Research Institute.