A histological method for quantifying Plasmodium falciparum in the brain in fatal paediatric cerebral malaria

Abstract

Background: The sequestration of Plasmodium falciparum-infected erythrocytes in brain microvasculature through cytoadherence to endothelium, is the hallmark of the definitive diagnosis of cerebral malaria and plays a critical role in malaria pathogenesis. The complex pathophysiology, which leads each patient to the final outcome of cerebral malaria, is multifaceted and thus, metrics to delineate specific patterns within cerebral malaria are needed to further parse patients.

Methods: A method was developed for quantification utilizing counts of capillary contents (early-stage parasites, late-stage parasites and fibrin) from histological preparations of brain tissue after death, and compared it to the standard approach, in which the percentage of parasitized vessels in cross-section is determined.

Results: Within the initial cohort of 50 patients, two different observers agreed closely on the percentage of vessels parasitized, pigmented parasites and pigment globules (ICC = 0.795-0.970). Correlations between observers for correct diagnostic classification were high (Kendall's tau-b = 0.8779, Kappa = 0.8413). When these methods were applied prospectively to a second set of 50 autopsy samples, they revealed a heterogeneous distribution of sequestered parasites in the brain with pigmented parasites and pigment globules present in the cerebellum > cortex > brainstem. There was no difference in the distribution of early stages of parasites or in the percentage of vessels parasitized across the same sites. The second cohort of cases was also used to test a previously published classification and regression tree (CART) analysis; the quantitative data alone were able to accurately classify and distinguish cerebral malaria from non-cerebral malaria. Classification errors occurred within a subclassification of cerebral malaria (CM1 vs CM2). A repeat CART analysis for the second cohort generated slightly different classification rules with more accurate subclassification, although misclassification still occurred.

Conclusions: The traditional measure of parasite sequestration in falciparum malaria, the percentage of vessels parasitized, is the most reliable and consistent for the general diagnosis of cerebral malaria. Methods that involve quantitative measures of different life cycle stages are useful for distinguishing patterns within the cerebral malaria population; these subclassifications may be important for studies of disease pathogenesis and ancillary treatment.

Figure 1

Examples of capillary contents. Note that only the capillaries in Panels B, D, F, H, J, K, and L would be adequate for counting by our method; the other panels are illustrative of parasite elements but would not be used for counting. All images haematoxylin and eosin at 1,000× original magnification.

Figure 2

The distribution across 28 sequential cases of vessel sequestration (as measured by % of vessels parasitized) across six sites within the brain are shown as stacked bars (maximum = 600 represented 100% of vessels parasitized × 6 sites). Cases are ordered by mean % of vessels parasitized. The dotted horizontal line represents the overall cut off for “cerebral malaria” (23.3% × 6 sites = 139.8) and demonstrates that, for total brain quantification, there is a clear cut-off between controls (non-cerebral malaria, cases 1–13) and cases (cerebral malaria, cases 14–28).

Figure 3

The distribution across 50 sequential cases of vessel sequestration (as measured by % of vessels parasitized) across three sites within the brain are shown as stacked bars (maximum = 300 represented 100% of vessels parasitized × 3 sites). Cases are ordered by mean % of vessels parasitized. The dotted horizontal line represents the overall cut-off for “cerebral malaria” (23.3% × 3 sites = 69.9) and demonstrates two outliers as follows: (grey arrow) a case of coma of other cause with anaemia and no evidence of retinopathy whose parietal lobe measures was <23.3% but cerebellum was >23.3%; and (black arrow) a case of severe malaria anaemia whose parietal lobe measure was >23.3% but the child did not meet the clinical case definition of cerebral malaria.

Figure 4

A graphical representation of the distribution of diagnoses across parasite elements counted (early stage = unpigmented rings, late stage = pigmented trophozoites, pigment = total pigment globules) by medians demonstrating the predominance of pigment in CM2 and the primary finding of late stage parasites in most infections for the validation set (n = 41).

Figure 5

A graphical representation of the first node of the classification tree demonstrating that the resultant cut-offs (23.3% and 21%) were stable but moot as the percentage of vessels parasitized showed no overlap when only patients meeting the clinical case definition of cerebral malaria were analysed.

Figure 6

Study flow chart and classification results from CART analysis. The initial training and testing study (on left) represents the original data from [15]. The validation study, using 40 new cases, is shown on the right with the original classifications rules applied (upper results table) and a de novo set of classifications rules (based on the 40 new cases, lower results table). The first node (separating patients with any cerebral sequestration from patients without evidence of cerebral sequestration) remained stable (23.3% vs 21%) with no classification errors. The second node (separating patients with sequestration only from patients with sequestration and extravascular pathology including ring haemorrhages) also remained stable (58.5 pg/100 vessels vs 55.5 pg/100 vessels). A third node was required (<= or >94.9 UPP/100 vessels) to separate the remainder of these latter two categories; however there were, in total, four classification errors using this new model.