Red blood cell (RBC) alloimmunization is a clinically-significant problem, with alloantibodies detected in about 1-3% of general populations and in up to 50% of chronically-transfused groups. It is not known why some individuals form alloantibodies to RBC antigens while others do not. One hypothesis is that certain disease states or conditions may predispose an individual to alloantibody development. To elucidate whether specific medical conditions may be associated with 'responder' or 'non-responder' status, we assessed comorbidities in a large group of transfused patients separated into alloimmunized and control cohorts. Our approach to investigate linkages between alloimmunization and disease state involved a novel technique called 'network analysis,' which allows visualization of all potentially significant disease associations within a studied population. All male patients with at least one alloantibody detected on a type and screen over a 54 year period (1961-2015) were extracted from a hospital database. For this study, patients who developed only 1 alloantibody following RBC transfusion were defined as 'responders', while those with >1 antibody were defined as 'hyper-responders'. Non-responders were randomly-selected, transfused individuals without detectable antibodies. Comorbidities were recorded systematically using ICD-9 codes documented in the electronic medical record. Spearman's coefficients were calculated to determine the association between alloimmunization and disease states; pairs with highly significant correlations (p<0.01) were selected to construct the network. In the network, each disease is represented by a node, where node size is proportional to disease prevalence in the cohort. Edges (lines that connect nodes) represent a statistically significant, paired correlation. We assessed the relationship between connections and prevalence, and explored clusters of correlated diseases. The dataset used for network analysis included 166 alloimmunized patients and 60 non-alloimmunized, but transfused controls. Cases were matched by age (mean = 74.4 years for responders and 73.5 years for controls, p=0.71). Several diseases were more prevalent in the non-alloimmunized control cohort compared to patients who developed alloantibodies, including: arthritis, acute renal failure, colon cancer, myeloproliferative neoplasm (with no increase in blasts), gastrointestinal bleed, pulmonary hypertension, sepsis, and pneumonia. All other conditions had a similar prevalence in each cohort. Network analysis was used to assess for relationships between disease states in alloimmunized individuals. This subsequent alloimmunization network was comprised of 78 nodes (55 diseases, 23 clinical characteristics) and resulted in 113 associations. The strength of the associations is reflected by the thickness of each edge in the analysis. Seven motifs (or clusters) of highly associated diseases were identified. Among patients with alloantibodies, arthritis, acute myeloid leukemia/myelodysplastic syndrome (AML/MDS), acute myocardial infarction, and peritonitis were found to be more prevalent in the hyper-responder group (>1 antibody) compared to responders (1 antibody). Of note, while the prevalence of AML/MDS and peritonitis in the control and alloimmunized cohorts was similar, patients with these diseases were associated with hyper-responder status and the development of alloantibodies to the D, C, c, and E antigens. Comorbidity network analysis identified conditions associated with the likelihood of forming >1 alloantibody. This included an inflammatory disorder (peritonitis) and diseases traditionally requiring chronic transfusion support (AML/MDS). Preventative strategies (e.g., phenotypic antigen matching) may be considered to mitigate antibody development in patients with these disorders. Future studies examining variables not investigated in this study, such as the influence of transfusion burden on the association between disease state and antibody formation, are warranted.

Disclosures

No relevant conflicts of interest to declare.

Author notes

*

Asterisk with author names denotes non-ASH members.

Sign in via your Institution