Ethics clearance for this study was obtained from the University of the Witwatersrand Ethics Committee (HREC) (M1704129). The HREC waived patient consent as validation for ClearLLab 10C testing utilised remnant samples previously submitted for routine diagnostic testing. All flow cytometric data analysed were de-identified; data were therefore effectively anonymised, and no patient identifiers were used.

This prospective observational cohort study²⁹ was performed at a national referral flow cytometry laboratory at the Charlotte Maxeke Johannesburg Academic National Health Laboratory Service Flow Cytometry Laboratory, Johannesburg, South Africa. The unit is an academic testing facility with a 300-sample monthly workload, referred from four large sister academic hospitals, other district-level facilities, and other regional centres across the national network. In addition, the laboratory participates in the United Kingdom National External Quality Assessment Scheme (Sheffield, United Kingdom) proficiency testing scheme for leukaemia immunophenotyping (Part 1) and Leukaemia Diagnostic Interpretation (Part 2).³⁰

Bodily specimens, including bone marrow aspirates, peripheral blood, cerebrospinal and pleural fluid, and lymph node tissue, referred to the National Health Laboratory Service in Johannesburg, South Africa, for leukaemia immunophenotyping by local sister hospitals were included in the study. After diagnostic testing with the laboratory developed test (LDT) was completed, the same sample was tested using the index method (ClearLLab 10C), provided that sufficient prepared cell concentrate for at least 4 aliquots of concentrated sample, or at least 1 mL of the whole sample, remained. All specimens for ClearLLab 10C evaluation were selected per the Bethesda guidelines.² The guideline explicitly outlines the appropriate use of flow cytometry for patients with clinically suspected oncological haematolymphoid neoplasms or patients with previously diagnosed oncological haematolymphoid disease enrolled for treatment. Samples with no pertinent history or clinical suspicion of haematological disease, older than 24 h, visibly haemolysed or clotted, with insufficient volume at receipt of the sample, or insufficient remnant sample to undertake index testing, were excluded from the study. One hundred and one remnant specimens (from 36 children and 65 adults from any ethnic or racial background) were processed for ClearLLab 10C from August through November 2019: 82 bone marrow aspirate samples and 15 peripheral blood samples were collected into dipotassium ethylenediaminetetraacetic acid. Three bodily fluid samples (two cerebrospinal and one pleural fluid) were submitted; a single lymph node biopsy was collected into saline.

For this study, the existing predicate method is the LDT; samples were analysed on the Becton Dickinson (BD) FACSCalibur (San Hose, California, United States). The index method is the ClearLLab 10C panel; samples were analysed on a Beckman Navios flow cytometer (BC, Miami, Florida, United States).

All sample cell concentrates were prepared for flow cytometric analysis according to current local standard operating procedures, irrespective of reagents used. Two to four 500 µL aliquots of blood or bone marrow per patient were prepared depending on the initial white blood cell count. The aliquots of the sample were placed in conical centrifuge tubes to which 14.5 mL of 0.899% solution of ammonium chloride (NH₄CL) containing 0.084% sodium bicarbonate and 0.0037% ethylenediaminetetraacetic acid (Merck, Darmstadt, Germany) was added to lyse the red blood cells. This was allowed to stand for 15 min, followed by a short 3-min spin at 3000 rpm. A further 5-min incubation time in the NH₄CL solution, followed by a 3-min spin, was performed if the red cells were macroscopically visible on the cell pellet. Samples were subsequently washed three times with a commercial phosphate-buffered saline (PBS) solution at pH 7.3 ± 2 (Oxoid LTD, Basingstoke, United Kingdom) containing 0.09% sodium azide (Merck, Darmstadt, Germany) and 0.2% bovine serum albumin (Biowest, Nuaille, France). Cell pellets were resuspended with 1 mL PBS. White cell count was obtained for each sample using 100 µL of the cell concentrate and 100 µL of BC FlowCount™^31,32 beads (BC, Miami, Florida, United States) in a 1 mL PBS solution on the Navios flow cytometer. The volume of cell concentrate estimated to contain ~1 × 10⁶ cells per aliquot was calculated, and this predetermined cell concentrate volume was added to each labelled marker tube of the LDT panel first. Remnant cell concentrates, or cell concentrates harvested from whole remnant samples, were aliquoted into the ClearLLab tubes. Each sample aliquot was incubated in the dark at room temperature (22 °C) for 15 min, followed by a final wash in PBS. The cell pellet was reconstituted with 700 µL of PBS and immediately run on either the FACSCalibur flow cytometer (BD Biosciences, San Jose, California, United States) for the LDT or the Navios flow cytometer (BC, Miami, Florida, United States) for the ClearLLab samples. Whole samples (remaining after cell concentrate preparation) were stored on the bench at room temperature (~22 °C) during testing; after analysis was completed, remnants of both the whole sample and the prepared samples were refrigerated at 4 °C.

Immunophenotypes were established using appropriately applied, pathologist-defined, sub-population gates that were unique to each case analysed. For the LDT data, a combination of light scatter and CD45 or CD19 gating was used to identify target cell (neoplastic) populations. For the ClearLLab 10C, initial gating included isolation of singlets followed by primary gating on CD45 positive events, with secondary gating applied to the identified target population. The specific antigen expression was documented for both the haematological neoplasm present as well as the normal haematological populations in the background (as an internal control of marker expression), based on the expression of all the available markers (antigens) across the four panels included in the ClearLLab 10C panel set. The specific presence or absence of antigens, and the respective specific intensity of staining of each antigen, was used to establish the diagnosis of specific subtypes of leukaemia or lymphoma per the World Health Organization’s classification of haemopoietic and lymphoid tissues^7,8. Also, the clinical history, other laboratory parameters (e.g. full blood count and white blood cells differential counts), and bone marrow aspirate morphology were taken into account.

Data were collected and collated into Microsoft Excel (Redmond, Washington, United States) spreadsheets for data analysis. Sensitivity outcomes were calculated; the clinical accuracy of the qualitative flow cytometric results (concordance of diagnosis as normal or abnormal) between the LDT and ClearLLab 10C outcomes were assessed for samples tested, and a contingency sensitivity (as the percentage of patients with malignancy) and specificity (the percentage of patients with no malignancy) table was constructed. In addition, the positive predictive value, as the probability of malignant outcomes with an abnormal immunophenotype detected, as well as negative predictive value as the probability of non-malignant outcomes in a patient with no disease, were also calculated. Thereafter, the ClearLLab results of each patient were compared to their clinicopathological or haematological diagnosis based on a similar approach by Hedley et al.²⁸ to demonstrate the objectivity of a comprehensive ClearLLab 10C diagnostic approach. Clinical concordance was achieved using the categories of ‘haematologically malignant’ or ‘haematologically non-malignant’, as suggested by the presence or absence of an abnormal immunophenotype. In this comparative analysis, both false positives and negatives, and true positives and negatives, as well as both positive predictive value and negative predictive value outcomes, were reported using ClearLLab outcomes as the objective standard.

The daily quality control results over the study period confirmed that daily background counts were consistently < 100 events in 100 s, carryover < 1% and Flow Check Pro results acceptable with HPCV < 2 for FL1 to FL5, HPCV < 3 for FL6 to FL8 and < 4 for FL9 and FL10 for the days that the patient samples were tested; Flow-Set Pro also confirmed acceptable fluorescence stability, with %CVs < 5% for the blue and violet laser (FL1–FL5 and FL9–FL10) and < 8% for the red laser (FL6–FL8). Further weekly CD4 reproducibility studies on the Navios confirmed that the flow count rate and absolute counts had a %CV < 5% for the duration of the validation. Precision and accuracy of ClearLLab controls were in agreement with the manufacturer’s claims for repeatability and accuracy of ClearLLab control cells (i.e. control cells assay values fell within the package insert limits and %CVs were consistently < 10%).

One hundred and one bodily specimens from 36 children and 65 adults, irrespective of ethnic or racial background, were enrolled into the study. These included: 82 bone marrow aspirates, 15 peripheral blood samples, one pleural and two cerebrospinal fluids, and one lymph node biopsy. Sixty seven of the 101 cases tested had haematological malignancy, while 34 cases had no evidence of malignancy despite a clinical suspicion (Figure 1). Among three bodily fluid samples tested, two cerebrospinal fluid samples showed evidence of tumour infiltration, one with T-cell acute lymphoblastic leukaemia, and another infiltrated with acute myeloid leukaemia; a separate (third) pleural fluid sample showed infiltration by follicular lymphoma, confirmed with CD10 expression and other markers suggesting B-cell maturity. The lymph node biopsy had a large T-cell lymphoproliferative disorder; additional LDT testing revealed an anaplastic large cell lymphoma (confirmed with CD30 expression). A diagnostic concordance of 100% was noted between the ClearLLab outcomes and the LDT, revealing identical diagnostic outcomes (Figure 2). These findings exceeded the manufacturer acceptance limits of > 70% sensitivity and >80% specificity for the ClearLLab testing system. All markers met the manufacturer’s specifications and claims, specifically those not typically included in first-line LDT testing, including CD16, CD38, CD56, CD64, CD123, CD200 and TCRγδ. Cases tested and reviewed using ClearLLab 10C were shown to have essentially similar immunophenotypes observed with the use of the LDT, leading to the same overall diagnostic outcome. In certain categories, however, diagnostic specificity matched diagnostic outcomes exactly, that is, there was 100% concordance of measured markers. Such examples include CD19/CD5 dual positivity for diagnosis of chronic lymphocytic leukaemia or CD19/CD10+ dual positivity for a common acute lymphoblastic leukaemia. Although overall diagnostic concordance was 100%, the four compact ClearLLab panels offered timely simultaneous, collated, and concise interpretation for both tumour and normal background populations. This improved efficiency was difficult to achieve, during interpretation and analysis of the predicate LDT, where limited 2–4-colour marker combinations had to be interpreted across in at least 10–12, but often as many as 20, separate panels.

The overall diagnostic sensitivity and specificity were comparable for the outcomes reported on corresponding clinicopathological bone marrow aspirate or trephine findings (Figure 3). The ClearLLab 10C system reliably identified and excluded all diseases and, in two cases, detected the presence of a B-cell lymphoproliferative neoplasm that was not detected morphologically on the corresponding bone marrow aspirate.

Firstly, while there was 100% concordance between the LDT and ClearLLab outcomes, only 101 case reviews were included. Secondly, although the ClearLLab instrument auto-setup makes managing daily quality control and review of complicated multicolour analyses much quicker, easier and standardised, the auto-setup itself still requires training of operators. Attending pathologists also need training and are required to familiarise themselves with Kaluza software and patterns of normal and aberrant haematological population maturation. Thirdly, the ClearLLab 10C M1, M2, T and B-cell panels allow for reasonably extensive discovery of most leukaemias, but there are currently no supplementary ‘second-line’ ClearLLab investigative panels available^7,8 to discern markers needed for full diagnoses; for example, cytoplasmic studies to discern myeloperoxidase or nuclear terminal deoxynucleotidyl transferase (TdT) expression, CD41, CD42b and CD61 for platelet marker investigation, or extended B-cell lymphoproliferative workup, including amongst others CD11c, CD23, CD30, CD43, CD79b and CD123 amongst others, is still performed in our laboratory to enable the World Health Organization classification of haematological disease. Some commercially produced, pre-dispensed, fixed panels marketed for in vitro rare event use^{27,38,43,44,45,46,47,48} could fill the gap. Such products could potentially be modified to extend the diagnostic repertoire of the ClearLLab 10C and provide for multicolour second-line immunophenotypic investigation, especially for B-cell lymphoproliferative disorders and plasma cell dyscrasias, but this needs further study.

In this study, there was excellent concordance between our LDT and the multicolour BC ClearLLab 10C panels, with 100% sensitivity and specificity recorded against existing LDT methods. There was also excellent concordance of ClearLLab 10C reporting to documented clinicopathological outcomes. The ClearLLab 10C panels, with manufacturer-standardised setup for colour compensation, appropriate quality control and data acquisition on a Navios flow cytometer, provides concise and comprehensive multi-immunophenotyping for typical leukaemias and lymphomas encountered during routine service delivery while substantially simplifying and standardising sample preparation and data acquisition.