Early diagnosis and treatment are key to malaria control. Currently malaria diagnosis in the field largely relies on microscopy and RDT, however both methods have limitations. Furthermore, malaria parasites are evolving to evade diagnosis, as demonstrated by the deletions of the pfhrp2 and pfhrp3 genes that allow parasites to escape detection by RDT [24]. Thus innovative tools are needed to overcome existing challenges toward global malaria elimination. In this study, a novel malaria diagnostic system, the Sysmex XN-31p was compared with conventional diagnostic methods in the Lake Victoria region, Kenya, and its applicability for use in the field was evaluated. Several studies have shown the capability of the XN-31p and its predecessor, the XN-30, to diagnose Plasmodium infections in different settings [12,13,14], but this is the first report showing the concordance of results between venous and capillary blood samples without any pre-treatment or sample dilution prior to analysis. Also this study shows that parasitaemia results obtained from capillary blood were stable even after 24 h, especially if the samples can be kept at chilled conditions. These positive findings unlock the potential use of the XN-31p as a diagnostic system for both clinical use and mass blood surveys in remote settings.
The diagnostic performance of XN-31p is quite comparable to that of existing field diagnostics, namely microscopy and RDT. The XN-31p was highly sensitive (1.000) and specific (0.986–1.000) against these conventional methods. These results are in good agreement with previous reports [12,13,14]. More importantly the performance was reproducible using capillary blood samples without any pre-treatment or dilution.
Detection of low-density Plasmodium infections by PCR is common in research but remains limited in clinical settings. Relative to PCR, the XN-31p detected fewer Plasmodium infections, resulting in a lower sensitivity (0.818–0.857) but maintaining a very high specificity (1.000). The lower diagnostic performance of the XN-31p is a disadvantage, however when compared to PCR, the XN-31p is easier to operate and provides results in a shorter period of time, making the XN-31p more suitable in clinical settings.
The XN-31p has three measurement modes. Whole blood (WB) and LM modes do not require any pre-treatment prior to the measurement, while PD mode needs sample dilution before the measurement. PD mode requires only 20 μL of blood and thus is useful when the amount of available sample is very limited. The applicability of capillary blood in PD mode was previously tested, however our own preliminary tests indicated low consistency in parasite counts and CBCs due to errors introduced during sample dilution. LM mode uses 130 μL of blood, which can be readily obtained through finger prick. Our comparison showed excellent concordance between venous and capillary blood samples for malaria diagnosis and species differentiation when samples were measured in LM mode. Good concordance was also observed for %iRBC, with the limits of agreement between −0.036 and 0.029%, which are well accepted in the clinical context. However, the parasite count per μL had fixed and proportional biases and the ranges for the limits of agreement were quite wide. Since there was no significant difference in RBC counts between capillary and venous blood samples (as shown in Fig. 3), the discrepancy must have derived from actual differences in iRBC count. The cause of differences in iRBC count is under investigation.
There was a high correlation but low concordance between the parasitaemia determined by the XN-31p and microscopy. By microscopy, parasite density was determined by counting the number of iRBC against 200 WBCs in the thick blood smear and assuming 8,000 WBCs per μL of blood. In this study, WBC counts were quite variable among participants (Fig. 3A), which might have contributed to the discordance. However, %iRBC was obtained without any assumption, thus the gap observed here may be purely due to the difference of two measurements. Similar trends in the concordance of parasitaemia between the XN-31p and microscopy were reported previously [14]. Although microscopy is considered the gold standard in malaria diagnosis, accurate determination of parasitaemia depends on a number of factors including the skills of microscopists and the quality of the stained blood smears, thus other quantitative methods such as qPCR may give insights into the accuracy of parasite count with XN-31p.
Malaria-positive samples are flagged by the XN-31p as “Malaria?(P.f.)”, “Malaria?(Others)” or “Malaria?(UNC)” based on the pattern of scattergram. Unfortunately, few non-P. falciparum malaria infections were included in this study, making it difficult to evaluate the ability of XN-31p to differentiate Plasmodium species. All P. ovale infections in this study were detected by PCR only and thus were likely to have low parasitaemia. It is reported that P. ovale infections tend to have lower parasitaemia because of the species preference to invade reticulocytes [25]. PCR detected one case of P. falciparum-P. malariae co-infection, which was not detected by other methods likely due to low parasitaemia. It would be interesting to test the diagnostic capability of the XN-31p in mixed-species infections since the scattergram patterns of RBC infected by different Plasmodium species may overlap with one another. No Plasmodium vivax was observed in this study, consistent with a previous report from the same area [26].
A small number of samples in our study had inconclusive results, reported as MI-RBC Abn Scattergram by the XN-31p. In the case of capillary blood, a previous study attributed this scattergram to the presence of Howell-Jolly body or crystalized Sickle haemoglobin (HbS) [12]. However in this study, HbS was reported in only 3 of 11 samples with MI-RBC Abn Scattergram, and the one sample reporting the presence of Howell-Jolly body did not report MI-RBC Abn Scattergram. Preliminary tests revealed a relatively high number of results reporting MI-RBC Abn Scattergram, which were likely caused by the finger prick method to obtain capillary blood. Repeated application of pressure on the finger can cause mechanical damage or deformation to blood cells, and damaged cells form aggregates with platelet and fibrin. The observation that inconclusive results were reported in more than twice as many capillary blood (11) as venous blood (5) samples is consistent with this hypothesis. To reduce the appearance of MI-RBC Abn Scattergram in this study, all laboratory technicians were retrained to follow the blood sampling procedure recommended by the Clinical and Laboratory Standards Institute (CLSI) (Additional file 1: Fig. S2). Interestingly most of the MI-RBC Abn Scattergram (80% for venous blood samples and 81.8% for capillary blood samples) were observed in PCR negative samples in this study. Further tests with a larger sample size may provide a clearer answer to the causes of MI-RBC Abn Scattergram.
Operations of the XN-31p require stable electricity supply that are often unavailable in health facilities providing services in malaria endemic areas. To extend the utility of XN-31p as a malaria diagnostic tool in remote communities, capillary blood samples kept at either room or chilled temperatures after 24 h were remeasured to simulate the time and condition of blood sample transport from distant villages to Homa Bay County Referral Hospital. Generally, samples stored at 2–8 °C had better concordance in parasitaemia with their 0-h results than samples stored at room temperature, with narrower limits of agreement. The concordance would be higher if one outlier were excluded, though the numbers of Plasmodium positive samples available for analyses were small. CBC data also showed better concordance in samples stored at chilled temperatures. Samples were kept cold in an inexpensive, food-grade insulated cooler box with ice packs that can be readily maintained in basic health facilities. Some studies showed the correlation between the distance from health facility and malaria prevalence [27,28,29,30]. The hub and spoke model described in this study can be exploited to advance universal access to malaria diagnosis and treatment.
One of the unique features of the XN-31p is the simultaneous detection of Plasmodium infection and measurement of CBCs, which may be useful in malaria eliminating areas. In such settings, lower malaria prevalence means fewer chances for microscopists to examine blood films and maintain competence in malaria diagnosis. Moreover, differential diagnosis of febrile diseases becomes more important in eliminating settings where malaria is no longer the main cause of fever [31,32,33,34]. The XN-31p can improve fever case management by excluding malaria as the cause of febrile illnesses. Combined with differential WBC counts, the XN-31p can further help clinical practitioners in rural health facilities to prescribe appropriate treatments.
This is the first test of the XN-31p in East Africa. The Lake Victoria basin has some of the highest malaria burden in Kenya, and previous cross-sectional surveys in Homa Bay County showed malaria prevalence of approximately 20% and 40% by microscopy and PCR, respectively [26, 35]. Malaria prevalence in the study area has decreased substantially due partly to the roll out of the indoor residual spraying (IRS) programme. Also the onset of the COVID-19 pandemic coincided with the start of this study, greatly reducing the number of outpatients seeking treatment at the hospital and resulting in a study sample size much smaller than anticipated due to altered care-seeking behaviours driven by social distancing measures such as lock downs and travel restrictions. In this study site, almost all outpatients presenting fever and/or malaria-related symptoms at health facilities are routinely tested for malaria. Thus many non-malarial cases were recruited for this study; only about 10% of the study participants were malaria positive. However, the results from this study are in general agreement with those from previous reports and provide evidence to support further development and evaluation of the XN-31p as a malaria diagnostic tool.
In summary, this study showed that in lieu of venous blood, capillary blood can be used directly without any pre-treatment or dilution for malaria diagnosis on the XN-31p automated haematology analyser. Moreover, capillary blood can be stored at chilled temperature for up to 24 h without adversely affecting malaria diagnostic and CBC results, broadening the appeal to utilise the XN-31p in a hub and spoke model as a rapid and accurate malaria diagnostic method for mass surveys and case confirmation in remote locations. The ability of the XN-31p to detect asymptomatic and submicroscopic infections and mixed-species infections needs to be investigated.