The value of bone marrow examinations performed in the investigation of HIV infected patients with cytopenias

Bone marrow examination (BME) is a reliable and effective tool in the diagnosis of many haematological and non‐haematological diseases and may be used to investigate unexplained cytopenia in human immunodeficiency virus (HIV) infected patients. The objective of this study was to determine the diagnoses made, diagnostic yield and unique diagnostic yield of BMEs performed to investigate cytopenias in HIV infected patients.

The mechanisms by which HIV infection results in cytopenias are complex and not fully elucidated. HIV may infect bone marrow (BM) stromal cells, T-lymphocytes and macrophages resulting in the dysregulation of BM cytokines and chemokines which in turn disturbs haematopoiesis. 3 Additionally, drug therapy and a predisposition to opportunistic infections, autoimmune diseases, haemophagocytic lymphohistiocytois (HLH) and certain malignancies are indirect ways in which HIV can affect haematopoiesis or cause peripheral destruction of blood cells. [3][4][5] The proposal that HIV can directly infect haematopoietic stem and progenitor cells is contested. [6][7][8][9] Bone marrow (BM) aspirate and trephine biopsy together with adjunctive investigations, herein referred to as BM examination (BME), may be utilised to investigate HIV-infected patients with unexplained cytopenia. [10][11][12][13][14][15][16][17][18] International studies show that the diagnostic yield (the percentage of cases in which a diagnosis is made) of BME in HIV-infected adults ranges from 19 to 47%. [10][11][12][13][14][15][16][17][18] Where the main or only indication for BME was cytopenia in an HIV-infected person, diagnostic yields of 14.8 to 30.5% were reported. 10,11,[14][15][16] Unique diagnostic yields for BME in patients infected with HIV range from 24-33%. [10][11] A unique diagnosis refers to one that was made by BME only. In some cases, alternate investigations can establish a diagnosis, less invasively or faster. 10,12,19,20 For instance, although BME is the gold standard for the assessment of iron stores, it is rarely indicated for the diagnosis of any haematinic deficiency. [21][22] While BME is a low-risk procedure, complications of variable severity have been reported. 23 Thus the necessity of BME has been questioned, particularly for disseminated opportunistic infections where a diagnosis could arguably be made by blood culture instead.
However, prior studies have demonstrated that the advantages of BME in the diagnosis of infection include the fact that histopathological examination can provide a diagnosis faster than either blood or BM culture and that there are instances where BM culture yields a diagnosis when blood culture is negative. Furthermore, the best diagnostic yield is achieved by the combination of blood culture with BME and BM culture. 12,19,20,24 South African studies from both the antiretroviral (ART) and the pre-ART eras, show that BM infection by opportunistic organisms is a common finding in HIV infected patients presenting with cytopenias and/or fever. In these studies, opportunistic infection was diagnosed by BME in 27.2-34.6% and in the majority, TB was the causative organism. [10][11][12] Morai et al, demonstrated that HIV infected patients receiving ART are less likely to have a diagnosis of BM infection than those not receiving ART (diagnosis of infection of 2% vs 16% respectively; p = 0.003). 13 In September 2016, all PLHIV in South Africa, regardless of clinical stage of disease and CD4 count, became eligible to receive ART. 25 While still only 71% of HIV-infected adult South Africans were receiving ART in 2019, increased availability and usage of ART may have impacted on the prevalence of opportunistic infections and other diseases affecting PLHIV. 1 Factors that are predictive of a diagnostic BME in HIV-infected patients have been explored. Leucopenia of <4 x 10 9 /L, absolute CD4 count of <100 cells/μL, the presence of wasting or oral thrush, and the coexistence of fever and cytopenia were significantly associated with a diagnostic BME. 10,16 Leucopenia of ≤4 x 10 9 /L, absolute neutrophil count of <0,5 x 10 9 /L, haemoglobin of <6 g/dL and previous TB were significantly associated with the establishment of a unique diagnosis. 11,12 Not receiving ART, leucopenia ≤4 x 10 9 /L, haematocrit of <25% and absolute CD4 count of ≤50 cells/μL were significantly associated with a diagnosis of infection. 12,13,20 Identification of further predictive factors may aid in the selection of patients who are likely to benefit from BME.

| Aims
This aim of this study was to assess the role of BM aspirate and trephine biopsy together with adjunctive investigations in finding a cause for unexplained cytopenia in HIV-infected patients by calculating the diagnostic yield and unique diagnostic yield of BME, describing the pathology found by BME and investigating for positive predictive factors of a diagnostic BME in this population.

| Study design and setting
This retrospective cross-sectional descriptive study was performed at Tygerberg Hospital, a tertiary academic hospital, and main teaching hospital for Stellenbosch University Faculty of Medicine and Health Sciences in Cape Town, South Africa. Records from the beginning of January 2015 until the end of May 2020 were used.

| Study population
BMEs performed for the main indication of unexplained cytopenia in HIV-infected adult patients 16 years of age and older, were included.
Exclusion criteria included BMEs done to further investigate a known haematological malignancy or monitor progression of a known disease, absence of an adequate quality core biopsy and BME samples received from referral hospitals.

| Data collection
BMEs that met study criteria were identified on the National Health Laboratory Service's Laboratory Information System (LIS) database.
The following information was extracted from the LIS: indication for and findings of BME, demographic information (age and sex), clinical features and comorbidities as stated by the treating clinician, full blood count (FBC), differential leucocyte count, direct antiglobulin test (DAT), haematinic studies (serum vitamin B12, folate and ferritin), HIV viral load, absolute CD4 count and adjunctive tests such as BM or peripheral blood flow cytometry, BM or peripheral blood culture and cytogenetic tests.
BM aspirate and trephine biopsy samples were taken from the posterior iliac crest in all cases. BM aspirate samples were stained using the May-Grünwald-Giemsa stain. The Prussian Blue (Perl's) stain was used to stain for storage iron on BM aspirate. Trephine biopsy specimens underwent fixation, decalcification, processing and embedding in paraffin wax. The haematoxylin and eosin stain was universally used.
Special stains for the investigation of infection were performed at the discretion of the pathologist. BM culture (bacterial, fungal or mycobacterial), cytogenetic tests and flow cytometry were performed at the request of the treating clinician or at the discretion of the pathologist.
Anaemia was defined as a haemoglobin of <13 g/dL in men and < 12 g/dL in women, neutropenia as an absolute neutrophil count of <1.5 x 10 9 /L and thrombocytopenia as a platelet count of <150 x 10 9 /L. 2 Indication for BME was categorised as pancytopenia, bicytopenia or single cytopenia depending on the number of cell lines reduced. Neutropenia rather than leucopenia was used to categorise cytopenia. 2 If administration of a red cell transfusion was suspected, the presumed pretransfusion haemoglobin result was recorded.
Results of absolute CD4 count, HIV viral load, serum B12 and ferritin performed up to 6 months prior to BME were recorded. Serum folate results from up to 1 month prior to BME were recorded.
A specific diagnosis was defined as any diagnosis made on BME.
A unique diagnosis was defined as a diagnosis made only on BME. Criteria for the diagnoses of malignancies were based on the latest available edition of the World Health Organisation's Classification of

Tumours of Haematopoietic and Lymphoid Tissues. Non-diagnostic
BMEs were defined as those in which criteria for a diagnosis was not met, although there may be bone marrow morphological changes present. For numerical parameters, the median and interquartile range were reported in most cases on account of skewed distribution of data. Age was the only parameter with a normal distribution, thus, the mean was reported. Fisher's exact test and Pearson Chi-square test were used to analyse the association between variables. A p value <0.05 was considered statistically significant.

| Ethical considerations
Ethical approval for the study was granted by the Stellenbosch University's Health Research Ethics Committee (reference number: S19/10/243).

| RESULTS
In total, 129 BMEs were performed on HIV-infected patients with the main indication of unexplained cytopenia. One BME was excluded as HIV status could not be verified. Four patients had two BMEs each during the study period. Thus 128 BMEs, performed on 124 patients, met the inclusion criteria.
Patient characteristics and selected investigations performed prior to BME are summarised in Table 1. The frequency and grading of the severity of cytopenias is summarised in Table 2 and an analysis of FBC and differential leucocyte counts is summarised in Table 3. Therapy refractory thrombocytopenia was the main indication for BME in 12 cases (9.4%) of which 5 were also anaemic and 2 had neutropenia. Immune thrombocytopenic purpura (ITP) was diagnosed in 9 of these cases (7%). BM involvement by a non-haematological malignancy was diagnosed in 1 case (0.8%) which had mild anaemia without leucoerythroblastosis or dacrocytes.

| Findings of bone marrow examination and adjunctive investigations
Increased lymphocytes (interstitial and in aggregates) were reported in 41 cases (32%) and were considered reactive in 36 cases (28%). A mature lymphoid neoplasm was diagnosed in 5 cases (3.9%).

| Diagnostic yield, unique diagnostic yield and summary of specific diagnoses made
The specific diagnoses made are summarised in Table 4. Two diagnoses were made in a single case (ITP and IDA), hence a total of 42 diagnoses were made in 41 cases. The overall diagnostic yield of BME was 32% (41/128). All diagnoses made were unique except for three diagnoses of IDA. The unique diagnostic yield was 30.5% (39/128).

| Positive predictive factors of a diagnostic bone marrow examination
Selected patient characteristics and haematological parameters were explored as predictors of a diagnostic BME (Table 5). None were found to be statistically significant.

| DISCUSSION
The overall diagnostic yield of 32% in this study is comparable to other studies which encompassed all indications for BME in HIV-infected patients. The diagnostic yields ranged from 19% to 47% in those studies. [10][11][12][13][14][15][16][17][18] Our diagnostic yield was slightly higher than the range of 14.8% to 30.5% reported where the main or only indication was cytopenia in an HIV-infected patient. 10  T A B L E 3 Summary of full blood count and leucocyte differential. This is likely due to their inclusion of only those patients admitted to the infectious diseases ward, thereby selecting a more immunocompromised population.
Possibly the most noteworthy finding of our study is that in contrast to most similar studies where opportunistic infection of the BM was the most common diagnosis, it was the least common diagnosis made in our study. [10][11][12] In Patients on ART are significantly less likely to have a diagnosis of infection on BME than those not on ART. 13 Compared to similar South African studies, our study population had the highest number of patients on ART and the highest median CD4 counts. [10][11][12] This suggests that our population was less immunocompromised and therefore, less susceptible to disseminated opportunistic infection. This is supported by the fact that granuloma formation was less common in our study with a prevalence of 11.7% versus 17.4% to 44% in other South African studies. [10][11][12] Infection, particularly TB, is the most common cause for BM granuloma formation in TB prevalent areas. 26 Thus if granuloma formation is used as a surrogate for BM infection, it could be concluded that BM infection was indeed less common in our population. Furthermore, absolute CD4 count is known to correlate with quality of granuloma formation with lower absolute CD4 counts, particularly <50 cells/μL, associated with poor granuloma formation. 27 As our study population had a much higher median absolute CD4 count, there ought to be well formed BM granulomata in response to infection if present.
However, inadequate investigation may have also contributed to our lower diagnostic yield of infection. In our study special stains for infection were performed in only 17.2% of cases and BM was cultured in only 22.7% of cases as compared to other studies where special stains for infection were performed in 55.7-100% and BM was cultured for TB or other organisms in 69.1-84% of cases. [10][11][12] The most common diagnosis made was PRCA in contrast to other studies where PRCA was one of the least common diagnoses. This is possibly because BME is routinely performed to confirm suspected diagnoses of PRCA at our centre but not at others. Final diagnoses of HL and BL were made in 2 cases and HL was also part of the differential diagnosis in an additional 2 cases. HIV is associated with an increased risk of high-grade B-cell lymphomas including BL, 5 and immunodeficiency-associated BL is by far the most common BL variant in adults in our population. 29 22 In patients with chronic kidney disease or heart failure, iron deficiency is possible with serum ferritin levels as high as 500 ug/L. Even higher levels of serum ferritin have been described in patients with depleted BM iron stores. 22 The cut-off serum ferritin

| Potential clinical value of BME in non-diagnostic cases
Despite the absence of a specific diagnosis by BME in 68% of cases, BME still provided useful information mainly by excluding serious BM pathology. In the majority of these cases BM cellularity was normal or hypercellular thereby indicating that a peripheral cause for the cytopenias should be sought. In certain cases, BME revealed pathology that was not diagnostic but rather required further investigation or monitoring with repeat BME. Non-specific HIV-associated BM abnormalities such as reactive lymphoid aggregates, an interstitial increase in reactive lymphocytes, increased polyclonal plasma cells, haemophagocytosis, dysplasia and granulomas were prevalent in non-diagnostic BMEs as well as in the study population overall. This is in keeping with the findings of previous studies of BM abnormalities in HIV. 33 This study did not identify any significant positive predictive factors of a diagnostic BME. This is possibly because of our relatively low number of cases. However, BME appears to be a reasonable investigation in cases of persistent cytopenia without an obvious explanation in HIV-infected individuals regardless of their CD4 count or viral load.

| Limitations
Data was restricted to what was accessible on the LIS and clinician request forms. As a result of insufficient information we could not investigate the positive predictive value of clinical features such as fever, oral thrush and wasting.

| CONCLUSION
This study proves that BME remains a useful diagnostic tool in HIV-infected patients with unexplained cytopenias. BME is a useful tool in the diagnosis of disseminated infection but special stains for infection and BM culture are under-utilised in our setting. In cases of presumed ITP that is therapy refractory or that has atypical clinicopathological findings, a BME is strongly recommended. Lastly, HIV-infected patients with cytopenias should undergo less invasive investigations to exclude haematinic deficiencies, sepsis and haemolysis on an individualised basis, prior to BME.

REPRODUCED MATERIAL
This article does not contain material reproduced from other sources.

AUTHOR CONTRIBUTIONS
CC devised the project and the main conceptual ideas, carried out data collection and curation and wrote and edited the manuscript. ZC and EM supervised all aspects of the project including conceptualisation, data analysis and reviewing and editing of the manuscript.

FUNDING INFORMATION
No funding was received for this manuscript.
T A B L E 5 Association between selected patient characteristics, haematological parameters and a diagnostic bone marrow examination