This was a retrospective record review of bone marrow aspirate and trephine examinations performed on adults admitted to the infectious disease (ID) ward of the Charlotte Maxeke Johannesburg Academic Hospital (CMJAH), in Johannesburg, SA, from 01 January 2012 to 31 December 2014.

Any patients, irrespective of their suspected diagnosis and HIV status, who had a BME in the ID ward during the study period were included in the study. There were no exclusion criteria. Patients admitted to the ID ward included any patient aged ≥ 16 years with a suspected ID that may or may not have potentially required isolation during their hospital stay, examples of which were patients with suspected tuberculosis, meningitis and opportunistic infections related to HIV infection.

The results of the bone marrow aspirate and trephine studies were obtained from the National Health Laboratory Services’ (NHLS) database. Data were extracted from these reports and transferred to a standardised data collection sheet. Each record was allocated a number and identifying features were removed. Only laboratory tests performed at the time of or within a month before or after the BME were used for this analysis.

The parameters documented were the patients’ age, gender, HIV status, CD4 and viral load (VL) as well as pre-BME blood test results, namely the full blood count, reticulocyte production index, and vitamin B12, ferritin and folate levels. Cytopenias were recorded. The indications for the BME as well as the final clinical diagnoses were documented. The diagnosis of MTB on BME was made through a positive bone marrow culture and/or positive MTB polymerase chain reaction (PCR) test. At the time of the study direct PCR testing of bone marrow for MTB was not routine practice; where such data was available it has been included in the study. The diagnosis of MAC was made on a positive bone marrow culture. The presence of acid-fast bacilli on Ziehl–Neelsen (ZN) stain and/or that of granuloma on trephine examination suggested a likely mycobacterial infection. However, the microbiological diagnosis remained unconfirmed unless marrow culture or simultaneous tissue or fluid sampling elsewhere in the patient provided positive identification of the pathogen.

If relevant clinical information or data were not available on the NHLS database, the individual patient’s hospital records were obtained and reviewed. Sputum for geneXpert MTB/RIF was routinely sent on all ID ward patients with a productive cough. Where indicated, the following tests were also performed: lumbar puncture, pleural tap or fine needle aspirate. A BME was performed if these investigations failed to provide a diagnosis. The results of all investigations were obtained from the NHLS database.

For the purposes of this study, the term ‘unique diagnosis’ refers to any diagnosis made on the BME that was not made with any other diagnostic tests or if the BME provided an answer more timeously than alternative tests. In particular, the latter refers to bone marrow cultures that flagged positive before other specimens. These ‘unique’ results were important for patient care and may have influenced subsequent outcomes.

Statistica, version 13, and Stata were used to analyse the data. Descriptive statistics were used for patient demographics. Means and standard deviations were used for haematological parameters. The Student’s t-test was used for comparison between continuous data with a normal distribution. The Mann–Whitney-U test was used in comparisons between data without a normal distribution, specifically in the comparison between possible predictive variables and unique diagnoses. A statistically significant result was defined as a p-value of < 0.05. Odds ratios (OR) were calculated for predictive determinants of a unique diagnosis on BME.

Ethics approval was obtained from the University of the Witwatersrand Human Research Ethics Committee (clearance number M150847).

A total of 327 bone marrow aspirate and trephine examinations were carried out in the adult ID ward during the study period. The study population consisted of 162 (49.5%) males and 165 (50.5%) females. The mean age of the study population was 36 years with a range of 17–65 years.

Overall, 314 patients (96%) were HIV-seropositive and 12 HIV-seronegative (3.7%). One patient’s HIV status was unknown. Amongst those with HIV infection, the median CD4 cell count was 47 cells/mm³ (1 cells/mm³ – 1069 cells/mm³) and 271 patients (86.3%) had a CD4 cell count of ≤ 200 cells/mm³. There were 128 patients on ART (40.8%) at the time of the bone marrow investigation.

The major indication for performing a BME was for the investigation of a peripheral blood cytopenia (54% of cases). Thirty-two per cent of BMEs were performed due to a clinical suspicion of MTB and 9% were performed owing to a suspicion of a malignant lesion with bone marrow involvement. Other indications, accounting for 5% of BMEs performed, were for the investigation of a thrombocytosis, leucocytosis or suspected disseminated cryptococcal infection.

Of the 327 BMEs performed, 158 (48.3%) yielded positive results (Table 1). Three of these cases yielded two diagnoses each. The most common findings were the presence of granulomata on trephine examination (57 cases, 17.4%) and a positive bone marrow culture for MTB (50 cases, 15.3%). The presence of granulomata together with a positive ZN stain on BME was found in 32 cases (9.8%).

In 77 cases (23.5% of the study population), the diagnoses made were unique to the BME (Table 2). Confirmed MTB was the unique diagnosis obtained in 17 of the 77 cases (22.1%) with unique diagnoses and confirmed MAC comprising three of the unique diagnoses obtained (2.6%). In two of these cases, MAC was diagnosed considerably faster on BME than peripheral blood culture (4 vs. 38 days and 8 vs. 26 days, respectively). Five patients had the diagnosis of a pure red cell aplasia made on BME. Four cases of haematological malignancies were identified. These comprised two cases of non-Hodgkin’s lymphoma (NHL), one case of Hodgkin’s lymphoma and one case of peripheral T cell lymphoma. Aplastic anaemia and disseminated cryptococcal disease were two additional unique diagnoses made on BME.

Three patients had two unique diagnoses obtained on bone marrow investigation. The patient with Hodgkin’s lymphoma also had the finding of granulomata on BME and was on empiric TB treatment. One patient with a positive Parvovirus B19 PCR slide also had a positive TB bone marrow culture. This patient was also on empiric TB treatment at the time of the BME. The third patient, who was not on TB treatment at the time of the BME, had both the diagnosis of NHL as well as a positive ZN stain together with granulomata observed on BME.

Of the laboratory parameters measured, a white cell count (WCC) ≤ 4 × 10⁹/L was found to be a significant predictor of a unique diagnosis on BME (OR: 2.38 and 95% confidence interval [CI] 1.37–4.14) (p = 0.002) (Table 3). A neutrophil count ≤ 0.5 × 10⁹/L was found more commonly in those with a unique diagnosis (p = 0.05); however, the numbers in this group were too small to interpret correctly.

Of the 327 bone marrow investigations performed, microbiological confirmation of MTB and MAC were made in 54 and 4 cases, respectively. Of the 226 bone marrow mycobacterial cultures performed 55 yielded positive results (24.3%). Of the 182 ZN stains performed 51 yielded positive results (28.0%). Six MTB PCRs were commented on as positive; however, it was not noted how many MTB PCR tests were performed that yielded negative results, and thus comparisons could not be performed. A WCC ≤ 4 × 10⁹/L and a CD4 cell count ≤ 50 cells/mm³ were found to be significant predictors of a microbiologically confirmed diagnosis of a mycobacterial infection (including MTB and MAC) on BME (Table 4).

There were seven patients with the diagnosis of MAC in the study population. In three of these patients, the BME was not diagnostic and MAC was found on a positive peripheral blood culture alone. Of the remaining four patients with MAC in the study population, one cultured MAC on bone marrow culture alone, namely a diagnosis unique to the BME. Amongst the other three patients, both the bone marrow and peripheral blood cultures were positive for MAC. In two of these cases, MAC was cultured faster on bone marrow than peripheral blood culture (unique diagnoses) and in one case MAC was cultured faster on peripheral blood hence not a unique diagnosis. This resulted in three of the four cases of MAC diagnosed on bone marrow culture that were unique to the BME. In all the cases of MAC, the CD4 cell counts were ≤ 35 cells/mm³. Four of the five patients who were on ART at the time had virological failure and one of the five presented as a suspected immune reconstitution syndrome after having been on ART for 3 months.

Positive results on BME were found in 48.3% of cases, which highlights the important persistent utility of BME in the investigation of patients with advanced immunosuppression and unexplained cytopenias, fever or the suspicion of disseminated mycobacterial infection. The finding of positive results in 48.3% of cases was marginally higher than previous studies in which diagnoses made on BME accounted for 25% – 47% of cases.^{3,4,5,8,9,10,12,13,14,21} In the current study, 23.6% of diagnoses made were unique to the BME. This finding was similar to the 24.9% unique diagnoses on BME found by Karstaedt et al.³; furthermore, this is also higher than international studies in which the prevalence of unique diagnoses has ranged between 8% and 10%.^3,13 Van Schalkyk et al.,⁴ however, had an even higher yield of 33% unique diagnoses in their cohort of 147 patients. The higher yield of unique diagnoses across South African studies may reflect the fact that in the developing world because of resource constraints this investigative modality is not widely available and, therefore, more consideration is given prior to performing a BME, thus those investigations that are performed are more likely to yield a positive result. Despite this, the study’s findings may be important to those practitioners in resource-constrained areas of SA as a guide as to which patients are likely to benefit from BME to aid with diagnosis.

The predominant unique diagnosis made on BME in the current study was that of MTB, which encompassed 22.1% (17/77) of BMEs with a unique diagnosis. This finding is in agreement with the findings by both Van Schalkwyk et al.⁴ and Karstaedt et al.,³ which also revealed MTB as the predominant unique diagnosis on BME. Brooke et al.⁵ in 1997 reviewed 215 bone marrow samples over a 9-year period in London and similarly found MTB as the predominant diagnosis (20% of all cases).

Other unique diagnoses found in the current study, which were similarly found in other studies, were disseminated MAC, haematological malignancies and disseminated cryptococcosis.^11,14,21 Most international studies, however, found the prevalence of MAC to be higher than that of MTB, reflecting differing disease profiles of mycobacterial infection in developing countries like SA in which MTB is predominant versus the developed countries in which MAC appears to be a more common diagnosis on BME.²²

This study found that a WCC ≤ 4 × 10⁹/L was a positive predictor of a unique diagnosis on BME (p = 0.002). A neutrophil count of ≤ 0.5 × 10⁹/L, although found more commonly in those with a unique diagnosis, was not found to be significant. Karstaedt et al.,³ similar to our study, found that a WCC < 4 × 10⁹/L predicted a unique diagnosis. Van Schalkwyk et al.⁴ found that a neutrophil count < 0.5 × 10⁹/L predicted a unique BME result; however, that study additionally found that a Hb < 6 g/dL was also a significant predictor of a unique diagnosis. More advanced HIV disease, with a lower CD4 cell count, was found to be an additional predictor of a positive BME in some international studies.^3,13 Keiser et al.¹³ and Luther et al.⁶ both found that a haematocrit < 25% and 30%, respectively, predicted positive results. This variable; however, was not measured in the current study.

Just over one-fifth of the unique diagnosis observed on BME resulted from confirmed MTB infection (22.1%). This finding highlights the persistent utility of the BMEs in diagnosing MTB despite the introduction of the Xpert MTB/RIF testing in early 2011.¹⁶ GeneXpert MTB/RIF testing of sputum has a reported sensitivity of 88% and a specificity of 99%.²³ Between 2012 and 2014, 690 435 Xpert MTB/RIF tests had been performed in Gauteng with the diagnosis of MTB made in 12% of cases.¹⁶ Despite access to this useful investigative modality being available at the CMJAH during the study period, we demonstrate that the BME still had utility to not only aid with diagnosis of MTB but also provide additional diagnoses that may not have been clinically suspected such as malignancies. An additional consideration is that many ill patients are unable to provide good quality sputum samples. Although induction of sputum may be attempted if this is unsuccessful, the availability of alternative diagnostic means such as BME is necessary.²⁴ In recent years, the urine lipoarabinomanin (urine LAM) test has become available as another modality to aid diagnosis of disseminated MTB infection; however, this testing modality was not available at our centre at the time of the study and its influence on the future necessity of BME is still to be determined.²⁵

There were a few potential limitations of this study. This was a single centre study undertaken in a single unit and so the results may not be generalisable to other units or to other hospitals in SA. Study patients were already in an ID ward with clinically advanced HIV at the time of the study, this at the outset would favour the presence of opportunistic infections, particularly MTB. Bone marrow findings were obtained from the retrospective review of published reports and individual bone marrow aspirate and trephine specimens were not reanalysed; we were, therefore, unable to control for any unidentified variables that may have influenced patient presentation. The subjectivity associated with the interpretation of the BMEs by the different assessors has to be noted. In addition, the fact that the study cohort was obtained specifically from an ID ward in which the yield of tuberculosis and the HIV sero-prevalence is anticipated to be higher than that of a general medical ward having patients with both communicable and non-communicable diseases may influence how the study findings may be applied to primary or district hospital settings in SA.