Recent years have seen the introduction of several new malaria diagnostic tests. Despite good sensitivity and specificity (1, 2), all malaria tests have the inherent disadvantage that they have to be requested explicitly, and lack of clinical suspicion has been reported as one of the main reasons for misdiagnosis, particularly in nonendemic countries (3). Other laboratory tests such as automated full blood counts (FBC) have not been very helpful in diagnosing malaria (4). Even one of the most frequently observed changes, thrombocytopenia, occurs in some 60% of patients with imported malaria (5), and it is associated with many other diseases.
A new generation of automated blood cell analyzers of the Cell-Dyn series (Abbott, Santa Clara, CA) use the so-called multiple-angle polarization scatter separation (MAPSS) technology to generate an FBC differential, which permits the detection of malaria pigment, or hemozoin, in monocytes (6, 7). These instruments differentiate eosinophils from neutrophils by their depolarizing granules, i.e., intensity of depolarized side scatter is interpreted as granularity. In the granularity-versus-lobularity plot, where lobularity corresponds to the total intensity of side scatter, eosinophils are depicted as green dots above a threshold line (Fig. 1a). The atypical appearance of monocytes as purple dots above the eosinophil and neutrophil threshold is indicative of phagocytized hemozoin (Fig. 1b) and caused by its birefringent (depolarizing) optical properties. In particular, the wavelength dependence of this phenomenon (8) is the subject of further investigation.
Figure 1. Granularity-versus-lobularity plot of a Cell-Dyn 3000 series full-blood count analyzer. a: Patient without malaria. b: Patient with Plasmodium falciparum malaria showing the typical appearance of hemozoin-containing monocytes in the eosinophil area. Note the abnormal eosinophil distribution, with some events at the top of the plot, most likely hemozoin-containing granulocytes, misclassified as eosinophils. Y axis (granularity), 90-degree depolarized light scatter; x axis (lobularity), 90-degree light scatter; green, eosinophils; blue, lymphocytes; purple, monocytes; orange, granulocytes.
Download figure to PowerPoint
After investigating 224 samples for malaria diagnosis, a study from South Africa found a sensitivity of 72% and specificity of 96% (9); and a study from Portugal found a sensitivity of 95% and a specificity of 88% in a series of 170 samples (10). In the South African study, a significantly lower sensitivity was observed in white Africans than in black Africans (9). This finding may suggest that one important factor is the number of previous infections encountered, as reflected in the individual's level of immunity against malaria, resulting in a lower sensitivity in cohorts of nonimmune individuals as encountered in nonendemic countries.
We evaluated the performance of the Cell-Dyn 3000 instrument to diagnose malaria in cases of imported malaria in Germany, to assess the relation between immune status and the instrument's performance, and to investigate the potential for possible improvements of the technique by comparing its capacity with a need-tailored experimental instrument using a MoFlo cell sorter (8).
- Top of page
- MATERIALS AND METHODS
- LITERATURE CITED
Overall results for microscopy and the Cell-Dyn 3000 are presented in Table 1. Of those 107 patients with malaria, 87 harbored P. falciparum, 13 had P. vivax, five had P. ovale, and two had P. malariae infections. There were no mixed infections. Mean age of patients was 37.1 years (median, 35.3 years; range, 6 months to 80 years); 231 (57.3%) were male and 172 were female (42.7%; male:female ratio, 1.3:1). Three hundred twenty-two (79.9%) patients were of European origin, whereas 81 (20.1%) were from malaria-endemic areas, predominantly from Africa (>99%).
Table 1. Overall Results of Microscopy and the Cell-Dyn 3000
Overall, Cell-Dyn 3000 failed to detect 55 of 107 malaria- positive samples (sensitivity 48.6%, specificity 96.2%). We observed no correlation between the number of PCMs and parasitemia levels. The mean number of PCMs in Cell-Dyn–positive patients was 5.5 (median, 2.0). For 40 of 55 patients (73%) with malaria but a negative Cell-Dyn result and for 40 of 52 patients (77%) with malaria and a positive Cell-Dyn result, material was available for IFAT performance. Cell-Dyn 3000 results of patients with malaria stratified by IFAT are shown in Table 2. Sensitivities to detect malaria were 73.7% for semi-immune and 28.6% for nonimmune patients. As can be seen from Figure 2, the probability to correctly identify malaria patients from depolarized side scatter with the Cell-Dyn increased with an increasing IFAT titer. Patients with a positive IFAT titer were identified as malaria positive by MAPSS in about 70% of cases. The existence of two groups of patients was confirmed by chi-square analysis of the data as given in Table 2, which yielded a significance level of P < 0.001. In addition, we tested for mean IFAT value equality in MAPSS-positive and -negative groups. At P < 0.0001, the two-sided Wilcoxon two-sample test for categorical data rejected equality in both groups.
Table 2. Cell-Dyn 3000 Results of Patients With Malaria Stratified by IFAT Resulta
Figure 2. Correlation of immune status with Cell-Dyn 3000 results. The number of observations are grouped for negative (<40) and positive (≥40) IFAT titer steps. P > 0.0001 in the two-sided Wilcoxon two-sample-test for differences between groups. CD, Cell-Dyn; IFAT, indirect fluorescent antibody test.
Download figure to PowerPoint
The large discrepancy in sensitivity of the Cell-Dyn 3000 analyzer for semi-immune (IFAT > 40) and nonimmune (IFAT < 40) patients is explained by the correlation between the relative concentration of PCMs as determined by the modified MoFlo flow cytometer and immune status. Figure 3 shows the comparison of the overall probability of malaria detection when using the routine instrument (Fig. 3a) and the normalized frequency of semi-immune and nonimmune patients (Fig. 3b) for PCM frequencies ranging from 10−6 to 10−2. Because the distribution of the normalized number of patients corresponding to semi-immune patients (median, 9 × 10−4) shifted toward the higher relative concentration of PCMs as opposed to the nonimmune patients (median, 1.5 × 10−4), sensitivity increased accordingly for semi-immune patients.
Figure 3. Correlation of Cell-Dyn 3000 results and immune status with relative PCM concentrations for 32 blood samples. a: Probability for malaria detection (black bars) versus relative concentration of PCMs. b: Normalized number of semi-immune (triangles) and nonimmune (dots) malaria patients for different relative concentrations of PCM. IFAT, indirect fluorescent antibody test; PCM, pigment-containing monocyte.
Download figure to PowerPoint
When employing the Cell-Dyn 3000 instrument, 11 false positives occurred, of which four episodes showed more than one PCM (2, 3, 9, and 13 PCMs, respectively). This may hint at a previous recent infection.
By counting up to 500,000 leukocytes with the modified MoFlo instrument and applying rare-event analysis, relative PCM concentrations as low as 3.25 × 10−5 were detectable. The total number of recorded events ranged from 150,000 to 500,000 and was adapted to the PCM abundance estimated during data acquisition. Relative frequencies of PCMs with respect to all leukocytes were determined for 32 blood samples of malaria-positive patients with the routine instrument and the modified flow cytometer. The correlation of relative concentrations determined with both instruments is illustrated in Figure 4. For the routine instrument, relative frequencies were calculated as the ratio of the number of PCMs and the number of leukocytes. Because the maximum number of analyzed events is limited to 5,000 leukocytes, a PCM count of one corresponds to a relative concentration of 2 × 10−4. This value represents the detection limit of the Cell-Dyn 3000 routine instrument. To include results for patients with zero PCM counts, these are represented by a relative frequency of 10−4, a value below the detection limit of 2 × 10−4. Error bars represent the statistical uncertainty only and were calculated as the square root of the number of recorded PCMs, assuming Poisson statistics. To estimate error bars for zero counts, we used the common practice in particle counting by taking PCM = 0 (+1/−0), i.e., all values between 0 and 1 are covered by the uncertainty of measurement. This was taken into account by choosing a threshold of 1.5 × 10−4 for the concentration of PCMs, as indicated by the horizontal line in Figure 4. It follows that with Cell-Dyn 3000 concentrations below the threshold (regions C and D) were measured for 10 blood samples. This results in false-negative classification of seven nonimmune and three semi-immune patients. For the modified MoFlo flow cytometer, detection efficiency to observe rare events was nominally improved by two orders of magnitude compared with the routine instrument. Relative concentrations as low as 2 × 10−6 were measured. However, the effective sensitivity to unambiguously identify PCM is lower because of interference with other events, e.g., degraded eosinophils. The threshold of 3.25 × 10−5 to discriminate malaria-positive from malaria-negative cases was derived by cross validation of a larger data set including 17 malaria-negative individuals (8). Compared with the routine instrument, the number of false-negative classifications is reduced from 10 to three (region C), leading to an improved overall sensitivity of 90.6% (100% in semi-immune and 83.3% in nonimmune patients).
Figure 4. Correlation between relative frequencies of pigment-containing monocytes determined with the routine analyzer (Cell-Dyn 3000) and the modified flow cytometer modified to detect rare events with high sensitivity. Thirty-two blood samples of semi-immune (triangles) and nonimmune (dots) malaria patients were studied. Thresholds used for malaria-positive and -negative classifications are represented as horizontal (routine analyzer) and vertical (high-sensitivity flow cytometer) dashed lines. Regions B and C, malaria-positive and -negative assignments with both instruments. Region D, Cell-Dyn-negative, MoFlo-positive malaria classification.
Download figure to PowerPoint
- Top of page
- MATERIALS AND METHODS
- LITERATURE CITED
The overall sensitivity reported here was much lower than that previously reported. With respect to the previous results (9, 10), we hypothesized that MAPSS analysis works best in semi-immune patients, because different populations have different amounts and kinetics of hemozoin-containing leukocytes (14, 15). Statistical analysis of our Berlin cohort results strongly supports our view. The most likely explanation is that semi-immune patients present later for pathophysiologic reasons because their “parasitic threshold” leads to more parasitic cycles and consequently to more pigment-containing macrophages before a delayed onset of symptoms, compared with nonimmune patients who become more rapidly symptomatic. This interpretation was encouraged by the correlation between immune status and relative concentration of PCMs.
In the two subgroups stratified by IFAT analysis, immunity status was positively correlated with an origin from endemic areas. These findings might help to explain the different sensitivities found for the subpopulations in a study from South Africa, with a much lower sensitivity in the white Africans (43%) than in the black Africans (90%). Semi-immune patients likely were represented more strongly in the black African than in the white African population.
Interestingly, the reported results are in contrast with those found in another study on imported malaria from Portugal in which a sensitivity of 95% was reported (10). However, demographic data and immune status were not analyzed in that study, and the semi-immune individuals might have been over-represented in the Portuguese study population. This seems even more plausible considering the close links between Portugal and its former colonies in Africa and, consequently, the large number of long-term residents from these malaria-endemic countries.
However, IFAT does measure a marker of antimalarial immunity that is not causal. For example, immigrants from endemic areas may present later than Europeans in the course of their illness due to non-medical reasons such as illegality. Although this possibility appears to be unlikely, we cannot exclude it from our data.
These findings may have important implications for the utility of this novel method to diagnose malaria in different settings. In endemic countries with a large number of semi-immune individuals, the sensitivity of the instrument will be high. However, frequent reinfections and low-grade parasitemia and persistence of hemozoin-containing monocytes in the circulation (14), even after successful treatment, may cause a low specificity. Conversely, in countries with malaria imported mainly by nonimmune individuals, patients may not have a sufficient number of hemozoin-containing monocytes to allow detection by the Cell-Dyn 3000 instrument when they seek medical attention. Therefore, with the current instrument's layout, a low sensitivity may be found in such a setting.
The overall specificity of 96.2% was similar to the one observed in the South African study (96%) but higher than that in the Portuguese study (88%). As discussed by Hänscheid et al. (10), the main reason for false-positive MAPSS results may be the persistence of hemozoin-containing monocytes after parasite clearance. In one study, pigment-containing white blood cells were still present in more than 70% of patients after parasite clearance, and the clearance time of PCMs has been reported to be up to 2 to 3 weeks (15, 16).
In the substudy comparing Cell-Dyn 3000 and MoFlo performances, the number of 10 patients falsely classified to be malaria negative with the routine instrument was reduced to three when using the flow cytometer with improved detection sensitivity. Compared with the routine instrument, the dynamic range of the high-sensitivity flow cytometer for rare-event detection is nominally extended by two orders of magnitude by increasing the total number of detected events accordingly. However, improvement in the detection limit of PCM amounts to one order of magnitude only, because at relative frequencies below 3.25 × 10−5 background events do not allow unambiguous identification of PCMs. Nonetheless, these results indicated that it might be possible to increase the malaria sensitivity of the Cell-Dyn instruments if more events than the “standard” 5,000 to 10,000 leukocytes are analyzed. To avoid a time-consuming FBC analysis in hematologic laboratories during routine use, modifications should include an option to analyze a larger quantity of cells in case of external, i.e., clinical, suspicion of malaria (query malaria option) and of internal suspicion, i.e., detection of a small quantity of PCMs during routine use, with activation of a built-in query malaria alert.
There is evidence that an atypical scattergram distribution of “eosinophils” indicating interspersed pigment-containing neutrophils is also a highly specific alternative to the granularity-versus-lobularity plot obtained from malaria patients (7). Taking this feature into account, the sensitivity of malaria detection in our mixed cohort of semi-immune and nonimmune patients would have risen to 57%. The full potential and limitations of this additional diagnostic feature of MAPSS have not been fully exploited and are subject to further studies. For example, differences in wavelength dependence of light scatter of pigment-containing leukocytes and other white blood cells might be used for an improved differentiation (8).
The number of PCMs reflects the number of pigment-carrying peripheral monocytes, which might be indicative of the total amount of malaria pigment present. Whereas a correlation of the number of PCM with parasitemia cannot be shown, the full potential of the method for correlation of MAPSS results with indication of progression towards severe disease, namely as a prognostic marker, awaits exploitation. In several studies from Africa and Asia, the amount of pigment found, particularly in neutrophils of malarious children, was a sensitive marker of progression toward and the prognosis of severe malaria in children (15–17).
In conclusion, the Cell-Dyn 3000 instrument has the potential to aid in the diagnosis of malaria. We have shown that the considerable variations of sensitivity and specificity might be caused by different ratios of semi-immune and nonimmune patients for the respective cohorts studied. Our results indicated that the method is particularly suited for semi-immune malaria patients and, hence, for diagnosis in malaria-endemic countries. In countries with imported malaria in mainly nonimmune individuals and with the instrument's current layout, i.e., wavelength and angles of observation chosen to detect light scatter and a limited number of collected events, sensitivity is too low to replace conventional diagnostic methods. Nonetheless, the advantage of this novel method is the fact that in these countries automated FBC are routine in the workup of febrile patients, and the instrument might allow detection of unsuspected cases in which clinical suspicion did not lead to a malaria-diagnostic test (6, 7, 18, 19). However, this technique has to be improved by allowing for an analysis of additional white blood cells in case of suspected malaria, by defining reliable on-screen identification criteria of hemozoin-containing neutrophils, and by exploiting the full potential of this technique to diagnose placental malaria and to facilitate a risk estimate for progression toward severe malaria. Cost reduction for purchase, supplies, and maintenance of these instruments would be an essential step toward permitting less-developed countries to exploit this technique.