The differential white cell count (DWCC) is a frequently requested laboratory investigation and is, historically, a labour-intensive test. Although the guidelines established by the international consensus group for haematology to identify samples where peripheral smear review may be omitted¹ have considerably reduced the workload on the morphology bench, manual smear review remains necessary in a substantial proportion of cases. With the number of skilled medical technologists on the decline worldwide, morphology skills are becoming ever scarcer, particularly in Africa where laboratory resources are, in general, grossly strained.^2,3 Novis et al. reported the smear review rate to be proportional to the number of occupied beds in the hospital served,⁴ which they speculated to reflect a higher pathology burden in larger hospitals. In sub-Saharan Africa, the pathology burden (and hence the need for smear review) is compounded by the HIV epidemic, which places an extra strain on haematology services because of the many haematological complications of HIV infection. There is thus a dire need for analysers that can improve laboratory efficiency in this setting.

The CellaVision™ DM96 (CellaVision AB, Lund, Sweden; hereafter, ‘DM96’) is a digital microscopy system that has the potential to minimise the time required by a morphologist to perform a manual DWCC. It comprises: an automated microscope that scans the blood smear; a digital camera that captures images of all the cellular and particulate material on the slide; and a computer that classifies each image by means of complex algorithms. It has been demonstrated to have good performance characteristics in developed countries,^5,6,7,8,9 but to date has not been evaluated in Africa, where the spectrum of pathology encountered is very different. Most notable in sub-Saharan Africa is the extremely high prevalence of HIV infection. More than 25.5 million people live with HIV in this region, compared with less than 2.5 million in Western/Central Europe and North America combined.¹⁰ The effect of this epidemic is greatest in communities with a poor socio-economic background. In South Africa, the most substantial impact is on state-sector hospitals. Because HIV infection is often associated with a number of morphological peculiarities of the white cells, including the frequent presence of atypical activated lymphocytes and abnormal nucleation of the neutrophils,¹¹ its impact on the performance of the DM96 is of interest for laboratories operating in areas with high HIV prevalence. The Chris Hani Baragwanath Academic Hospital (CHBAH) is a large referral centre in Johannesburg, South Africa, which serves a community with a poor socio-economic background and high HIV prevalence. The aim of this study was to assess the accuracy of the DWCC generated by the DM96 in the CHBAH laboratory, with emphasis on its performance in samples collected from HIV-positive patients.

Patient demographic and clinical data are summarised in Table 1. The median white cell count (WCC) was 6.76 × 10⁹/L (range 0.28–262). The WCC was < 1.5 × 10⁹/L in 14 patients (9.4%) and > 50 × 10⁹/L in 12 patients (8.1%). Slightly over half (n = 79; 53%) of patients were HIV-positive, whilst ~40% (n = 61) had a history of malignancy. The prevalence of malignancy was similar between HIV-positive and HIV-negative patients, although the spectrum of malignant disease varied substantially between these two groups. The dominant malignancy amongst HIV-positive patients was high-grade lymphomas, whereas leukaemias were more common amongst HIV-negative patients. Despite this difference, the proportion of patients with abnormal cells present in the peripheral blood was similar between the two groups. Not surprisingly, evidence of infection was substantially more common amongst HIV-positive patients.

In this study, we assessed the performance of the DM96 as compared to a manual DWCC in 149 samples collected from patients with a wide range of infections and haematological pathologies in a large South African state hospital serving a population with high HIV prevalence. Results from similar studies performed in developed countries have varied. Park et al. showed excellent correlation (CC > 0.9) for all cell types except promyelocytes and basophils.¹³ In contrast, although most other studies have also shown good correlations for neutrophils, lymphocytes and blasts, correlations were generally poorer for monocytes (CC 0.67–0.83), eosinophils (CC 0.73–0.88), and basophils (0.05–0.76).^6,7,8,9 Similarly in our study, there was excellent correlation (CC > 0.95) for both neutrophils and blasts and substantially weaker correlation (CC < 0.75) for eosinophils and basophils. Interestingly, the correlations for both lymphocytes (CC = 0.86) and monocytes (CC = 0.51) were poorer in our study than those reported previously. Nonetheless, we found that agreement was within acceptable limits of bias for both the monocyte and lymphocyte values, and accuracy was thus judged to be adequate for these parameters. The diversity of results between studies may be attributable, in part, to the proportion of pathological samples included in each study. In addition, Park et al. found that correlations improved in samples with low WCCs when the DM96 was preset to analyse a higher number of white cells (300–500), which may have impacted their overall correlation results.¹³ The relatively poorer correlation consistently seen for monocytes, basophils and/or eosinophils is likely a result of the predictably poor precision expected when cells present in small numbers are assessed by a limited cell count. Reassuringly, where discrepancies between the methods were evident in our study, the differences translated into small changes in absolute cell counts, the clinical significance of which was negligible.

Although HIV infection has well documented effects on white cell morphology,¹¹ misclassification rates in our study were not associated with HIV status. Misclassification rates were significantly associated with WCC and the presence of malignant cells in the peripheral blood, whereas chemotherapy exposure, a history of prior malignancy and the presence of infection were not. ‘Unidentified cells’ made up 3.5% of cells in our study, but overall, 16% required reclassification. Our misclassification rate was higher than that described previously by Rollins-Raval, Raval and Contis, who assessed the performance of the DM96 over a six-month period at three separate sites.¹⁴ The number of unidentified cells in that study was ~1% and the number of misclassified cells ranged from 4.6% – 12.7%. The higher misclassification rate in our study is likely a product of the large number of specimens containing malignant cells and selection of a large number of samples with extreme WCCs for the purpose of evaluating the analyser. Thus, samples with abnormal WCCs were disproportionately prevalent in comparison to the routine laboratory workload. The overall accuracy of the DM96 is therefore likely to be better than shown in our study, particularly when the WCC is near normal and malignant cells are not present. That being said, in the current era of sophisticated full blood count analysers able to perform accurate DWCCs in a timely manner, the necessity of performing manual DWCCs is restricted to samples in which the analyser fails to perform an accurate DWCC because of the presence of abnormal white cells. Consequently, the accuracy of results for samples with abnormal cells present is of greatest interest, as these specimens would potentially be eligible for analysis with the DM96 because of the need for a manual DWCC.

The time required to perform a DM96 DWCC has been demonstrated to be less than that required for a manual DWCC,^6,8,9 which raises hopes for greater laboratory efficiency when a manual DWCC is needed. However, the increased misclassification rate in the presence of abnormal cells shown in our study necessitates reclassification of a greater number of cells, which prolongs the time required to perform a DM96 DWCC in this setting. This could conceivably negate the marginal reduction in the time it takes for experienced morphologists to perform a DWCC using the DM96 as compared to a manual count. Cornet, Perol and Troussard suggested that the DM96 would prove to be time efficient, as the few unidentified cells could be quickly and easily classified and validated, thus reducing the time spent on microscopy by technical staff.⁵ However, we found that the number of unidentified cells comprised only about 20% of the misclassified cells; thus, re-assignment of only the unidentified cells would compromise accuracy to an unacceptable extent. Time efficiency is also undermined in samples with very low WCCs, where accuracy is reportedly increased by pre-setting the analyser to count 300–500 cells. However, this improvement comes at the expense of a longer analysis time.¹³ Moreover, the improvements in cell recognition and flagging technology in automated analysers mean that the vast majority of samples can be reviewed by scanning the peripheral smear without the need for a manual DWCC. Thus, in samples without a substantial number of abnormal cells present, the potential analytical time advantage expected from the DM96 is eliminated by the performance of a ‘smear scan’ in lieu of a manual DWCC. Given the benefits of a smear scan over a manual DWCC in the majority of samples, as well as the poorer time efficiency of the DM96 anticipated in samples with leukopenia or abnormal cells present, the utility of the DM96 is placed in question. This is emphasised by the user-dependent nature of correlation studies shown by Briggs et al., where the accuracy of the DM96 was noticeably poorer when the analysis was performed by less-experienced microscopists.⁶ Clearly, although the DM96 is a brilliant piece of innovative technology, it does not eliminate the need for skilled morphologists.