Frequency of benign neutropenia among Black versus White individuals undergoing a bone marrow assessment

Abstract Healthy individuals in the United States identified as having Black race have lower neutrophil counts, on average, than individuals identified as having White race, which could result in more negative diagnostic evaluations for neutropenia. To test this hypothesis, the proportion of evaluations where the final diagnosis was clinically insignificant neutropenia for Black and White individuals who underwent an evaluation by a haematologist that included a bone marrow (BM) biopsy to investigate neutropenia was assessed. 172 individuals without prior haematological diagnoses who underwent a haematological evaluation to investigate neutropenia. Individuals diagnosed with clinically insignificant neutropenia between Black and White individuals were compared using a propensity‐score‐adjusted logistic regression. Of 172 individuals, 42 (24%) were classified as Black race, 86 (50%) were males, and the 79 (46%) were over 18 years old. A BM biopsy did not identify pathology in 95% (40 of 42) of Black individuals and 68% (89 of 130) of White Individuals. Black individuals (25 of 42 [60%]) received a final diagnosis of clinically insignificant neutropenia, compared to White individuals (12 of 130 [9%]) (adjusted odds ratio =7.9, 95% CI: 3.1 – 21.1). We conclude that black individuals were more likely to receive a diagnosis of clinically insignificant neutropenia after haematological assessment.

genotype is strongly confounded with race, clinical decision-making related to the evaluation of neutropenia could differentially impact racial groups. 11 Lower baseline neutrophil counts could predispose Black individuals to undergo clinical evaluations, including invasive diagnostic procedures such as bone marrow biopsy (BMB), to identify a pathological cause contributing to the lower counts. Since the lower neutrophil counts among rs2814778-CC genotype carriers are not associated with underlying disease or other clinically meaningful sequelae, these investigations would not be expected to identify underlying pathology. 12 We hypothesized that, because of the high frequency of the rs2814778-CC genotype among individuals of Black race, this predisposition to lower neutrophils could result in a measurable race-related healthcare disparity. To test this hypothesis, we analysed a clinical data set to determine whether Black individuals who underwent an evaluation by a haematologist that included a BMB to investigate a finding of neutropenia were more likely than White individuals to be diagnosed with a clinically insignificant neutropenia.

| Study population
Individuals were identified using Vanderbilt University Medical Center's (VUMC) Synthetic Derivative (SD) resource, which is a deidentified data set comprising a large portion of data captured through the medical centre's electronic health record (EHR). 13 The SD database currently contains records for over 2.8 million individuals, with no defined exclusions. The study population was derived from a set of 20,353 individuals with a race recorded as either 'White' or 'Black' in the EHR, and a BMB pathology report in their clinical records. Individuals with an established diagnosis based on ICD-9/ICD-10 codes for a haematological cancer, a blood transfusion, an organ transplant or chemotherapy/radiation therapy appearing 3 or more days prior to their first biopsy were excluded.
A keyword search was then used to identify all BMB reports that contained any of the phrases ('leukopenia', 'leucopenia', 'low white', 'decreased white', 'reduced white', 'low wbc', 'decreased wbc', 'reduced wbc', 'neutropenia', 'low neutro', 'decreased neutro', 'reduced neutro', 'lymphopenia', ' anc', 'anc', 'lymphopenia', 'granulopenia' and 'agranulocytosis') as part of the clinical indication for the biopsy. The clinical record of each individual was manually reviewed. As part of the manual review, only those individuals with a BMB performed at VUMC that included an indication of low neutrophils listed in their BMB report and a clinical evaluation by a haematologist were retained. Those with a prior diagnosis of haematological disease, a prior BMB or previously diagnosed cancer identified by either automated screening using ICD-9/ICD-10 codes or manual review were excluded.

This study was evaluated by the VUMC Institutional Review
Board and determined to be non-human subjects research. 13

| Clinical data
Race, gender, as recorded in the EHR, and Duffy antigen testing results were extracted from structured data tables. Age was based on the date of the first BMB report. White blood cell count, absolute neutrophil count (ANC), platelet count and haemoglobin levels were extracted from the haematology clinical note preceding the BMB. All haematology notes prior the BMB were reviewed and findings of concern to the haematologist (in addition to neutropenia) were coded including haematological abnormalities (anaemia, thrombocytopenia, cell type elevations, presence of blast cells, thrombosis, gamma globulin disorders and other), comorbidities (rheumatological disease, splenomegaly, trisomy 21 and other) and symptoms (fever, concerns for recurrent infections, bone pain, lymphadenopathy, night sweats and unintentional weight loss, and other). Presence or absence of benign neutropenia in the stated differential diagnosis was recorded. For each individual, the first BMB hematopathology report, which comprised histology, flow cytometry, immunochemistry and/or other relevant studies was reviewed.
All haematology notes after the BMB were reviewed to identify the final diagnosis explaining the individual's neutropenia. An individual was assigned a diagnosis of 'clinically insignificant neutropenia' if the final diagnosis was 'benign [ethnic] neutropenia' or neutropenia of no clear aetiology and assessed to likely be of no clinical significance with respect to adverse health outcomes. If the final assessment included several possible diagnoses, the diagnoses was coded as 'undetermined'; and if the evaluation was not completed because the patient did not return for follow-up, the diagnosis was 'lost to follow-up'. All clinical data were reviewed by a physician blinded to the individual's race (JDM). Each hematopathology assessment was then categorized as either having a clinically significant haematological or cytogenetic abnormality, such as evidence of malignancy or myelodysplastic syndrome, myelofibrosis, maturation arrest, significant hypoplasia suggestive of aplastic anaemia, etc. as defined by a haematologist (SCB). Upon review and interpretation of the report, each BMB was categorized as normal or abnormal. Data were extracted to a REDCap database. 14,15 Bone marrow biopsies were performed that typically also include analysis of the aspirate as well as the decalcified biopsy. The p-value for the independent variable is reported.

| Analysis
The proportion of individuals with a diagnosis of clinically insignificant neutropenia was calculated separately for Black and White individuals and is presented for the overall study sample and stratified by select baseline characteristics. To test the independent association between clinically insignificant neutropenia and Black race, as compared to White race, multivariable logistic regression was used where the diagnosis of clinically insignificant neutropenia was the dependent variable and Black race as the independent variable. To control for factors associated with being of Black race, the regression model included a propensity-score for being in the Black race group. Weights for the propensity-score components were determined by logistic regression analysis using Black race as the dependent variable and the following independent variables: All analyses used 2-sided tests and, unless otherwise noted, a p-value < 0.05 was considered significant.

| Baseline characteristics
After exclusions, there were 172 individuals who had a haematological evaluation including a BMB for a clinical indication that included neutropenia and without a previously established haematological diagnosis ( Figure 1). Among these, 42 (24%) were classified as Black race and 86 (50%) were males ( Table 1) Overall, 49 (29%) individuals had no comorbidities or symptoms other than neutropenia (isolated neutropenia) that were of concern to the haematologist. The proportions of isolated neutropenia among were 41% for Black and 25% for White individuals. The respective proportions of Black individuals with any concerning haematological abnormalities, comorbidities, or symptoms were 31%, 14% and 29%, and for White individuals were 49%, 19% and 41% ( Table 1). Black individuals were more likely to have clinically insignificant neutropenia included among the differential diagnoses, as compared to White individuals (45.2% vs. 8.5%).

| Bone marrow evaluations
Black individuals were more likely than White individuals to have a bone marrow biopsy (often including flow cytometry and/or cytogenetic evaluation) that did not identify an abnormality. Forty

| Diagnosis by haematologist
Overall, 37 (22%) individuals were diagnosed with a clinically insignificant neutropenia. Baseline characteristics significantly associated with an increased likelihood of a clinically insignificant F I G U R E 1 Selection of the bone marrow biopsy population. The study population was drawn from individuals identified through a Vanderbilt University Medical Center de-identified electronic health record resource neutropenia included older age, higher ANC, higher haemoglobin, absence of a fever, and having isolated neutropenia (Table S2). In a multivariable logistic regression model adjusted for a propensity-score for factors associated with Black versus White race, Black race was associated with a significantly increased probability of having a final diagnosis of clinically insignificant neutropenia (odds ratio =7.9, 95% CI: 3.1 -21.1, p = 9 × 10 −7 ) (Table S3).

| DISCUSS ION
Approximately 60%-65% of African-Americans carry the Duffy Null phenotype that associates with lower neutrophil counts. The rs2814778-C single nucleotide polymorphism underlying this genotype confers the Duffy Null red blood cell antigen phenotype [Fy(ab-)]. [16][17][18] The lower ANC associated with the Duffy Null phenotype is  22 If ANC values remain stable, and in the absence of concerning symptoms (i.e., an isolated neutropenia), a bone marrow biopsy evaluation is typically not warranted.
We would advocate that assessing for the Duffy Null genotype (by genetic testing) or phenotype (by antigen detection) also be a standard part of the work-up for a WBC or ANC value that is concerning to a provider, prior to more extensive evaluation or serial monitoring, and regardless of race. 16  Black race treated at three medical centres who had a BMB to investigate an isolated low white blood cell count, 97% of individuals carried the rs2814778-CC (Duffy Null) genotype. 12 Furthermore, 97% of the biopsies among genotype carriers demonstrated benign findings. In the absence of knowing the Duffy phenotype status, there is uncertainty as to whether an observed low ANC value is expected for an individual, and this uncertainty may contribute to an extensive haematological work-up.
Often the Duffy phenotype is not measured clinically, but inferred based on an individual's race and observed ANC.
Consistently, in this study, only 1 Black individual had documented antigen testing, indicating that, even among haematologists, testing may not be a routine part of a clinical work-up. However, an inference that an individual carries the Duffy Null phenotype could be inaccurate over 40% of the time among African-Americans.
Furthermore, incorrectly assuming a Black individual has the Duffy Null phenotype could lead a provider to conclude that a low ANC is of no clinical significance, possibly resulting in a delayed diagnosis.
We suspect that many of the Black patients in this study received an intensive haematological work-up, including a BMB, because their neutrophil counts fell below an established clinical reference range that is not calibrated to their underlying genetic predisposition.
While an ANC<1500 cells/µl is the commonly accepted threshold for neutropenia, a recent study observed that up to 5% of Black individuals with the Duffy Null phenotype may have a value that falls below this threshold. 23 In contrast, less than 0.1% of Black individuals without the null phenotype fell below this value. Thus, Black individuals with this benign phenotype are considerably more likely to have a value that is considered to be clinically abnormal and are at increased risk of an unnecessary clinical work-up. In this study, 65% of Black individuals with an ANC <500 cells/µl were diagnosed with clinically insignificant neutropenia, suggesting that the phenotype may predispose to considerably lower ANC values. Separate clinical reference ranges for those with and without the Duffy Null phenotype would address this disparity.
We examined a clinical endpoint that included an evaluation by a haematologist in order to ensure that the final diagnosis of clinically insignificant neutropenia was accurate. However, the underlying problem of health care disparities among those with the Duffy Null phenotype manifests in other, and sometimes more subtle, contexts beyond repeated testing for low WBC levels. For instance, these individuals are more likely to have medications stopped or delayed because of a low ANC not recognized to be related to the Duffy Null phenotype. [24][25][26] They are also more likely to be excluded from clinical trials and other studies that have ANC thresholds for inclusion. 27,28 Again, many of these disparities could be attenuated with WBC and ANC reference ranges calibrated to the underlying biology.
It is important to note that, in the absence of isolated neutropenia, the clinical decision-making process surrounding neutropenia becomes more complex. Thus, when Black individuals with clinically insignificant neutropenia due to the Duffy Null phenotype present with additional symptoms such as anaemia, thrombocytopenia, fever, lymphadenopathy, weight loss, or fatigue, their low ANC may influence medical decision-making and these patients may undergo a more extensive diagnostic work-up to rule out more serious medical conditions. Abnormalities in two cell lines can be suggestive of an underlying bone marrow disorder. However, if the lower value for neutrophils is due to a benign predisposition, the likelihood of an underlying bone marrow predisposition is lower. Consistently, in this study, individuals of Black race, as compared to those of White race, were less likely to have an underlying bone marrow abnormality, even with a second haematological abnormality. Thus, their predisposition to a lower ANC is likely contributing to incorrect clinical inferences.
Further studies would be needed to ascertain whether genotypespecific references ranges would avoid unnecessary procedures.
The age distributions differed between the races. For White individuals, 37% were under 10 years old and 26% were over 50 years old. In contrast, for Black individuals, the respective percentages were 24% and 9%. Haematological malignancies are more common in early or late life, so the bimodal age distribution among White individuals corresponds to this expected disease pattern. In contrast, Black individuals were more likely to undergo evaluations in older childhood to middle life when leukaemia or other bone marrow diseases occur less frequently. Such a pattern would be expected if the haematological abnormality (neutropenia) of clinical concern was not driven by underlying disease.
This study has limitations. The data came from a single tertiary care medical centre and the study population and results may not be generalizable to other communities. Racial groups are comprised of heterogeneous collections of individuals representing a range of ancestries; binning individuals into two groups largely based on skin colour ignores important biological differences. 29 In addition, race assignment in the EHR may represent either self-reported race or race assigned by others. Data were derived from EHRs and reflect only the care obtained within this care system and do not include encounters from outside facilities. The bone marrow and final diagnoses represent the assessments of the pathologists and haematologists and the complete set of data used in these assessments were not available for re-review in this de-identified data resource. Black and White individuals differed with respect to several relevant baseline characteristics (age, ANC count, burden of comorbidities), which could reflect a different predisposition toward neutropenia or selection biases related to race and referral for evaluation.
In summary, this retrospective study found that Black individuals undergoing an evaluation by a haematologist that included a BM biopsy for neutropenia were more likely to be diagnosed with a clinically insignificant neutropenia of no clinical significance, as compared to White individuals. We speculate that this is due to the benign Duffy Null phenotype/genotype that is common among Black individuals and that is associated with lower neutrophil counts.
Establishing reference ranges based on the underlying this phenotype/genotype may avoid healthcare disparities attributable to the lower ANC counts in this group.

I R B A PPROVA L
This study was evaluated by the Vanderbilt University Medical Center (VUMC) Institutional Review Board prior to data collection and determined to be non-human subject's research.

DATA AVA I L A B I L I T Y S TAT E M E N T
Access to data from the VUMC resources can be requested through the investigators, and requires completing appropriate data use and sharing agreements and IRB approvals.