Automated detection of malaria pigment: feasibility for malaria diagnosing in an area with seasonal malaria in northern Namibia

Authors


Corresponding Author Dr Jeroen van Dillen, Department of Obstetrics and Gynaecology, HAGA Hospital, location Leyenburg, Leyweg 275, 2545CH The Hague, The Netherlands. Tel.: +31 70 362 3838; Fax: +31 70 359 2964; E-mail: gwamupanda@caiway.nl

Summary

Objective  To evaluate the feasibility of automated malaria detection with the Cell-Dyn® 3700 (Abbott Diagnostics, Santa Clara, CA, USA) haematology analyser for diagnosing malaria in northern Namibia.

Methods  From April to June 2003, all patients with a positive blood smear result and a subset of patients with no suspicion of malaria were included. Blood smear and a venous blood sample (to determine haemoglobin, platelet and malaria pigment levels) were collected from each patient. Malaria pigment test characteristics, correlations with blood parameters and pigment clearance time were calculated. Finally, a subset of blood samples was run twice to evaluate the consistency of test outcome.

Results  Two hundred and eight patients were included. Ninety had a positive blood smear result of which 84 tested positive for malaria pigment and 118 patients had a negative blood smear result of which four tested positive for malaria pigment. Test characteristics as compared with microscopy were as follows: sensitivity 0.93, specificity 0.97, positive predictive value 0.95, negative predictive value 0.95. Rerun of the blood samples resulted in a change of diagnosis in 14%. After 4 weeks, 33% of patients with an initially positive pigment result still tested positive. Malaria pigment was found to be negatively correlated with haemoglobin.

Conclusions  Automated detection of malaria pigment is a useful diagnostic tool in this semi-rural area. In low-risk malaria season, the test can be used for diagnosing malaria because of the high sensitivity. In high-risk malaria season, the test can be used for excluding malaria in case of a negative pigment result because of the high specificity.

Abstract

Objectif  évaluer la faisabilité de la détection automatisée de la malaria avec l'analyseur d'hématologie Cell-Dyn 3700, pour le diagnostic de la malaria dans le nord de la Namibie.

Méthodes  D'avril à juin 2003, tous les patients avec un frottis sanguin positif et un ensemble de patients sans suspicion de malaria ont été inclus dans l’étude. Pour chaque patient, un frottis de sang et un échantillon de sang veineux (pour la détermination des taux d'hémoglobine, du platelet et du pigment de la malaria) ont été collectés. Les caracteristiques du test du pigment de la malaria, les corrélations avec les paramètres sanguins et la clearance du pigment ont été calculés. Un certain nombre d’échantillons sanguins a été testé deux fois pour évaluer la consistance des résultats.

Résultats  Au total, 208 patient ont été testés. 90 patients avaient un frottis sanguin positif et parmi eux, 84 avaient un test positif pour le pigment de la malaria. 118 patients avaient un frottis sanguine négatif et parmi eux, 4 avaient un test positif pour le pigment de la malaria. Les caracteristiques du test comparés à la microscopie étaient les suivants: sensibilité 0,93%, spécificité 0,97%, valeur prédictive positive 0,95% et valeur prédictive négative 0,95%. 14% des échantillons testés deux fois avaient un résultat différent. Apres 4 semaines, 33% des patients avec un test positif pour le pigment de la malaria étaient toujours positifs. Une corrélation négative a été observée entre le pigment de la malaria et l'hémoglobine.

Conclusion  La détection automatisée du pigment de la malaria est un outil de diagnostic utile dans cette zone semi rurale. Au cours de la saison avec un risque faible pour la malaria, le test, grâce à sa haute spécificité, peut être utilisé pour exclure la malaria dans le cas d'un résultat négatif pour le pigment.

Abstract

Objetivo  Evaluar la viabilidad de un sistema de detección automatizada para el diagnóstico de malaria en Namibia del norte, utilizando el analizador de hematología Cell-Dyn 3700.

Métodos  Entre abril y junio del 2003, se incluyeron todos los pacientes con una lámina positiva y un grupo de pacientes sin sospecha de malaria. A cada paciente se le hizo una lámina y se le tomó una muestra de sangre venosa (para determinar hemoglobina y niveles de plaquetas y de pigmento de malaria). Se determinaron las características de la prueba para pigmento de malaria, su correlación con los parámetros sanguíneos y se calculó el tiempo de depuración del pigmento. Finalmente, se corrió dos veces un subgrupo de muestras de sangre, con el fin de evaluar la consistencia de los resultados.

Resultados  Se incluyeron 208 pacientes de los cuales 90 tenían una lámina positiva. De estos últimos, 84 fueron positivos para pigmento de malaria, mientras que de los 118 pacientes con una lámina negativa, 4 fueron positivos para pigmento de malaria. Las características de la prueba en comparación con la microscopía fueron las siguientes: sensibilidad 0.93, especificidad 0.97, valor predictivo positivo 0.95, valor predictivo negativo 0.95. El segundo análisis de las muestras de sangre resultó en un cambio de diagnóstico del 14%. Tras cuatro semanas, el 33% de los pacientes con un resultado positivo inicial para pigmento continuaban dando positivo. La detección automatizada del pigmento de malaria es una herramienta útil para el diagnóstico en esta área semi-rural. En la época de bajo riesgo de malaria, esta prueba puede utilizarse para diagnosticar la malaria gracias a su alta sensibilidad. En épocas de alto riesgo de malaria, y gracias a su alta especificidad, la prueba puede utilizarse para excluir malaria en caso de un resultado de pigmento negativo.

Introduction

One of the cornerstones of the global malaria control strategy is effective disease management by prompt and accurate diagnosis. However, the non-specific clinical presentation, high prevalence of asymptomatic carriers, lack of resources and insufficient access to trained healthcare workers all hinder diagnosing malaria (WHO 1999). Several approaches are being applied, of which clinical diagnosis is most widely used. Although it is inexpensive, quick and does not require special equipment, it is also unreliable: overdiagnosing and misuse of antimalarial drugs can be considerable (Reyburn et al. 2004). However, in areas with high rates of transmission and in resource-poor settings, treatment based on clinical diagnosis is considered justifiable (WHO 1999). Careful examination of a well-prepared and well-stained blood film by an expert microscopist remains the ‘gold standard’ for detecting and identifying malaria parasites (Makler et al. 1998). It is very sensitive, relatively cheap and also identifies Plasmodium species. However, it is labour intensive and requires special equipment and trained personnel. The optimal sensitivity (>90%) and specificity (100%) of this method therefore often change under field conditions with sensitivity ranging from 10 to 72% and specificity from 56% to 100% (Coleman et al. 2002; Reyburn et al. 2004). Antigen detection tests (rapid tests) use rapid immunochromatographic techniques to identify parasite antigens with reported sensitivity ranging from 88% to 93% and specificity from 92% to 96% (Cruciani et al. 2004). Untrained staff can do the test and no equipment is required. However, the costs involved are often too high for rural hospitals, the test cannot quantify the density of infection and results remain positive after (successful) treatment because of persistence of circulating antigen (Moody 2002).

Detection of haemozoin (or malaria pigment) by microscopy has been attempted both in erythrocytes and leucocytes. Haemozoin is a haem polymer resulting from the breakdown of haemoglobin by Plasmodium species (Sullivan & Meshnickb 1996). Several techniques were introduced, including dark-field staining and polarized light microscopy. Microscopically, haemozoin-carrying leucocytes were visible in >90% of patients with severe malaria and in 87% of children with mild malaria, while only 58% of adults with mild malaria could be diagnosed (Metzger et al. 1995; Day et al. 1996). Together with the time-consuming and complicated procedures of most staining methods (Lawrence & Olson 1986) and the disadvantages of microscopy under field conditions as mentioned earlier, this technique seems of limited value as single diagnostic tool, although it can assist in malaria diagnosis in ‘smear negative’ cases of (severe) malaria because of sequestration of parasites (Silamut et al. 1999).

Detection of haemozoin by Cell-Dyn automated haematology analysers was first reported in a South African study (Mendelow et al. 1999). Cell-Dyn haematology analysers, incorporating flow-cytometrical principles and being used for full blood counts, differentiate eosinophils from neutrophils by their granules, which depolarize light specifically. This characteristic can be plotted in a scatter diagram where eosinophils are presented as green dots above a threshold line and the other white blood cells below (Mendelow et al. 1999). Phagocytic monocytes and neutrophils, which ingested haemozoin also depolarize light (Metzger et al. 1995) and seem to do this in the same specific way as eosinophils. When samples of malaria infected patients are processed by the Cell-Dyn, purple dots (pigment dots) appear in the eosinophil area of the scatter diagram and this is caused by the presence of depolarizing haemozoin in phagocytic monocytes and neutrophils (Hanscheid et al. 2000). Several studies evaluated this method with sensitivity ranging from 52% to 95% and specificity from 72% to 100% (Mendelow et al. 1999; Hanscheid et al. 2001; Wever et al. 2002; Fawzi et al. 2003; Scott et al. 2003; Dromigny et al. 2005). However, most of these studies were carried out in non-endemic areas or with cases of imported malaria (Mendelow et al. 1999; Hanscheid et al. 2001; Fawzi et al. 2003; Scott et al. 2003). The aim of this study was to evaluate the feasibility of automated malaria detection with the Cell-Dyn 3700 haematology analyser for diagnosing malaria in Northern Namibia, an area with seasonal malaria following the rains.

Materials and methods

Study area

In Namibia, there are approximately 400 000 malaria cases and 500 deaths because of malaria each year and malaria is seen as one of the four major challenges for the Namibian Health System (El Obeid & Mendelsohn 2001). Malaria is predominantly found in the northern regions where it is seasonal following the rains. Plasmodium falciparum accounts for 97% of all malaria infections in Namibia (El Obeid & Mendelsohn 2001). The Ministry of Health and Social Services developed guidelines for the diagnosis and treatment of malaria in Namibia (Ministry of Health and Social Services Namibia 1995). According to these guidelines diagnosis should, if feasible, be confirmed by microscopic examination of a blood smear. However, because most patients are examined and treated at remote clinics, clinical diagnosis of malaria is the rule rather than the exception. Patients with treatment failures, possible chloroquine resistance or severe malaria should be transferred to the hospital where diagnosis can be confirmed or excluded by light microscopy (Ministry of Health and Social Services Namibia 1995).

Onandjokwe Lutheran Hospital is a 450-bed district and referral hospital situated in Northern Namibia with an estimated catchment population of 200–300 thousand people. The incidence of malaria is approximately 300 cases per 1000 people (Mendelsohn et al. 2000). Peak malaria transmission ranges from February to July during which period Onandjokwe District is graded as a high-risk malaria area. The laboratory services at the hospital are run by the privatized Namibia Institute of Pathology (NIP) and include services for parasitology, haematology and chemistry. The haematology unit uses a full blood count analyser (Cell-Dyn® 3700), which identifies haemozoin. Running costs are 1.2 USD for a haemoglobin result (which optionally includes a malaria pigment result) as compared with 3.7 USD for a malaria smear. The Cell-Dyn was introduced in the hospital in 2002 and in the same year a pilot study was performed in which patients who tested positive for malaria pigment and negative for blood smear were rechecked for parasites: of 322 samples, 76 were re-checked and an additional 43 patients tested positive for malaria by microscopy (from 246 to 289). The calculated sensitivity of microscopy was 85%.

Method

The objective of this study was threefold. First, the diagnostic value of malaria pigment was evaluated in a prospective clinical setting where patients with proven malaria (positive blood smear result) and patients with no suspicion of malaria were included. The Cell-Dyn was tested for diagnosing malaria in a population with a low pre-test likelihood (randomly chosen subset of patients presented at the department of general medicine with no clinical suspicion of malaria) and one with a ‘high pre-test likelihood’ (patients with a positive blood smear result). In addition, correlations between the pigment dot count and the parasite level, platelet count and haemoglobin level were analysed.

Second, all patients who tested positive for malaria pigment were followed for evaluation of the time pigment results remain positive after (successful) treatment. They were asked to come for follow-up on post-treatment days 3, 7, 14, 21 and 28. Follow-up was ended when a patient tested negative for malaria pigment or when the 28 days follow-up was completed.

Third, a subset of pigment positive samples (n = 105) was run twice to evaluate the consistency of the test with respect to dot count and test outcome: positive or negative.

Data collection and management

From each patient, the following data were recorded: blood was taken by finger prick for thick blood smear and stained with 10% Giemsa solution after which asexual parasites were counted against 200 white blood cells. Parasitaemia was calculated assuming a white blood cell density of 8000/mm3. A venous blood sample was run for haemoglobin, platelet and malaria pigment count using the Cell-Dyn 3700 haematology analyser. The number of pigment dots was counted and a result of one or more dots was regarded as a positive malaria pigment result. A blood specimen, obtained from the same finger prick as used for malaria smear, was taken to perform a locally available rapid manual test (Immuno-Mal P.F. Kenza Diagnostics, Kya Sand, South Africa). The Immuno-Mal is a rapid qualitative immunoassay for the determination of P. falciparum-specific histidine-rich protein II (PF HRP2). The test was performed in patients with a positive malaria pigment and negative blood smear result. In these cases, a positive rapid test would either indicate a successfully treated patient or a false negative smear and smears were cross-checked.

On follow-up days, the above-mentioned examinations were repeated. Only patients living within a radius of 10 km from the hospital were included and transport costs were compensated. Case record forms were anonymously taken and patients were only included after informed consent.

Data analysis

Data entry was performed on a weekly basis using Microsoft Excel. Analysis was performed using both Microsoft Excel and SPSS. To determine malaria pigment test characteristics, sensitivity, specificity, negative predictive value (NPV), positive predictive value (PPV) and likelihood ratios were calculated. A receiver operating characteristic (ROC)-curve was made to assess the best cut-off point for the number of pigment dots, regarded as a positive test result. Spearman rank correlation test was used to calculate possible correlations between the pigment dot count and the parasite level, platelet count and haemoglobin level. Fisher's exact test was used to calculate a possible correlation between a high-dot count and severe anaemia. A P-value of <0.05 was considered statistically significant.

Results

Cell-Dyn test characteristics for diagnosing malaria

From April to June 2003, 208 patients were included of which 90 had a positive blood smear result. Of 90 patients, 84 tested positive for malaria pigment on admission. Of the remaining six patients, three tested positive for malaria pigment on day 3, one still tested negative on day 3 and the other two were lost to follow-up. Of 118 patients with a negative blood smear result, four tested positive for malaria pigment. The rapid test in these four patients turned out to be positive in three cases (75%). Re-check of blood smears from these three patients did not reveal malaria parasites. Initial characteristics are presented in Table 1. Malaria pigment test characteristics as compared with microscopy were as follows: sensitivity 0.93, specificity 0.97, PPV 0.95, NPV 0.95, positive likelihood ratio 31, negative likelihood ratio 0.07. Sensitivity and specificity of the malaria pigment test can be altered with adjustment of the number of pigment dots making the result positive (cut-off point). The amount of pigment dots presented in the scatter diagrams ranged from 0 to 150 dots. An ROC curve indicated that a cut-off point of ≥1 dot results in the highest differentiating capacity (Figure 1). Malaria pigment was negatively correlated with haemoglobin (Spearman rank correlation: Rs = −0.379, P < 0.001) (Figure 2) and a dot count of >14 was correlated with severe anaemia (Hb < 7g/dl) (Fisher's exact test: P = 0.024, odds ratio = 0.119, 95% confidence interval = 0.015–0.978). Malaria pigment was not significantly correlated with parasitaemia and platelet level (Spearman rank correlation: Rs = −0.041, P = 0.705 and Rs = 0.157, P = 0.148, respectively).

Table 1.   Initial characteristics of the total study population (range or percentage)
 Malaria smear
Positive (n = 90)Negative (n = 118)
Age (years)23.8 (0.7–82)35.4 (13–99)
Sex (male)45 (50%)35 (30%)
Parasitaemia (parasites/mm3)27 786 (320–285 714)
Haemoglobin (g/l)9.9 (3.5–16.9)13.1 (5.9–22.1)
Platelet count (×109/l)104 (8.6–602)255 (24.2–523)
Malaria pigment (positive)84 (93.3%)4 (3.4%)
Figure 1.

 Receiver operating characteristic curve for the changing sensitivity and specificity according to the amount of malaria pigment dots regarded as a positive test result.

Figure 2.

 Scatter graph of the correlation between pigment dot count and haemoglobin.

Pigment clearance time

Of 84 patients with a positive malaria smear and positive malaria pigment result, 33 completed follow-up. Reasons for loss to follow-up included patient refusal (n = 8) and logistical reasons (n = 8). For 34 patients it remained unknown and one patient died because of cerebral malaria. The results are shown in Table 2. The mean follow-up for all five visits was 80%. Thirty-three per cent of the patients who completed follow-up still tested positive for malaria pigment after 4 weeks. If all patients lost to follow-up were regarded positive for malaria pigment (worst case), 62 (73%) would have tested positive for malaria pigment after 4 weeks, while in the best case scenario (all lost to follow-up assumed pigment negative), only 11 (13%) would.

Table 2.   Patients attending follow-up (% of expected) and pigment results (% of tested)
 Number tested (%)Positive pigment result (%)Negative pigment result (%)
  1. * The percentage of positive pigment results was calculated using both tested patients plus previously tested patients with a negative pigment result.

Day 084 (100)84 (100)
Day 358 (69)56 (97)2 (3)
Day 745 (80)42 (89)*3 (7)
Day 1435 (83)29 (73)*6 (17)
Day 2124 (83)14 (40)*10 (42)
Day 2812 (86)11 (33)*1 (8)

On day 0, the mean number of malaria pigment dots was 32 (range, 1–150) and it gradually decreased till day 14. On day 21, a moderate increase was seen (Figure 3). During follow-up, six patients were found to have either a re-infection or treatment failure and in four there was an increase in the amount of dots. However, not all patients with an increase were found to test positive at microscopy. When patients with a re-infection or treatment failure were excluded, the mean dot count continued to decrease till day 28 (Figure 3).

Figure 3.

 Mean number of malaria pigment dots during follow-up.

Consistency of pigment dots

A total of 105 samples were run twice (88 positive and 17 negative pigment results, including patients on follow-up days). The median difference was two dots (range, 0–37). If a cut-off point of ≥1 pigment dot(s) was regarded as a positive test result, the diagnosis changed in 15 patients (14.3%). For eight patients, the malaria pigment result changed from positive to negative and for seven from negative to positive. If a cut-off point of ≥2 dots was regarded as a positive test result, the diagnosis changed in 14 patients (13.3%; seven changed to a positive result) and with a cut-off point of ≥3 dots, the diagnosis changed in nine patients (8.6%; six changed to a positive result).

Discussion

Automated detection of malaria was found to be a valuable alternative for diagnosing malaria in this district hospital with a sensitivity of 93% and a specificity of 97%, as compared with microscopy. These results are consistent with earlier studies (Mendelow et al. 1999; Hanscheid et al. 2001; Scott et al. 2003; Dromigny et al. 2005). Apparently, the analyser is more sensitive than microscopy in detecting haemozoin-containing leucocytes, especially in patients with mild malaria. Using microscopy, the reported percentage of haemozoin-containing leucocytes on initial presentation is 83% (Metzger et al. 1995) (vs. 93% using automated detection).

In this semi-rural district hospital with seasonal malaria, microscopy (parasite detection) was found to be less sensitive (85%). This was based upon an earlier evaluation in 2002 (unpublished). Microscopy requires considerable expertise and other reports have shown a disappointing accuracy under field conditions, with errors in either direction (false-positives and false-negatives) (Coleman et al. 2002; Reyburn et al. 2004). For screening purposes, malaria pigment is therefore a better test than conventional microscopy in this hospital. Furthermore, the test is very simple, does not need any additional training for the laboratory technician and the running time is very short (Wever et al. 2002). In Portugal, it was shown that malaria pigment can assist the diagnostic process in patients with severe febrile illness, where a full blood count is often routinely performed and clinical suspicion of malaria might be low (Hanscheid et al. 2001). During the study period, three patients (not included in the study) tested positive for malaria pigment during routine blood analysis, while they were not suspected of having malaria. In all three cases, malaria was confirmed by microscopy.

However, disadvantages of automated malaria detection are the reported inability of quantification and differentiation of Plasmodium species and the prolonged time the pigment can remain positive (Hanscheid et al. 2000). Additionally, the initial and maintenance costs of an automated haematology analyser puts this test out of reach for most (rural) African hospitals. Especially the prolonged time pigment results remain positive limits its usefulness. In our study, 33% of patients still tested pigment positive after 4 weeks. In high-risk malaria areas, this could cause high rates of false-positive test results. Earlier studies reported clearance times up to 3 weeks (Hanscheid & Valadas 1999; Scott et al. 2003; Suh et al. 2003). However, these observations were based on patient history (Hanscheid & Valadas 1999) or Plasmodium vivax cases (Suh et al. 2003), not on prospective data. In contrast to P. falciparum, P. vivax detection by the Cell-Dyn largely depends on intraerythrocytic haemozoin, rather than haemozoin-containing leucocytes (Fawzi et al. 2003). As erythrocytes have half the lifespan of monocytes, this could explain the short-time pigment results remained positive after treatment in the P. vivax study (mean 3 days, range 3–14 days).

Two microscopy studies reported on this subject and described clearance times up to 21 days for haemozoin-containing monocytes (Metzger et al. 1995; Day et al. 1996). Automated analysis seems to be more sensitive in detecting haemozoin than microscopy and this might be the reason for the extended clearance times found in this study. Possible explanations are the long lifespan of (pigment-containing) monocytes and phagocytosis of circulating haemozoin by newly formed monocytes. Possible sources for continuous circulating haemozoin after parasite clearance are gametocytes, neutrophil remnants and the liver and spleen. Gametocytes are known to contain large amounts of pigment (Langreth et al. 1978) and can remain present for 16–24 days post-treatment (Smalley & Sinden 1977). Gametocyte density after antimalarial treatment depends on the duration of illness prior to presentation, type of drug (especially pyrimethamine–sulphadoxine is related to high levels of gametocytaemia post-treatment) and drug-resistance (Sowunmi & Fateye 2003). Haemozoin remains present in the liver and spleen for at least 9 months after parasite clearance (Levesque et al. 1999). It has been suggested that haemozoin-laden macrophages might gradually lyse with time and release haemozoin into the circulation (Levesque et al. 1999). A final possible source is treatment failure because of (partial) chloroquine-resistance. Chloroquine inhibits haemozoin production and in the case of chloroquine resistance, this inhibition is diminished. Because of sequestration (which is common in falciparum malaria) or a parasite load below the detection limit of microscopy, low levels of parasitaemia can result in a negative malaria smear, but continuing circulating haemozoin. Clearly, further investigation on this subject is needed. Insight into the causative factors might be helpful in improving the Cell-Dyn.

Four patients with a negative malaria smear tested positive for malaria pigment. Three of them had a positive rapid test and malaria smears were selected for a thorough re-check. However, the re-check did not reveal malaria parasites and the patients were regarded as successfully treated. Specificity of PF HRP2 rapid tests is known to be significantly reduced because of persistent antigenaemia for up to 10 days post-treatment (Mharakurwa & Shiff 1997). These four patients were attending our outpatient department for non-specific complaints and might have had a self-resolving malaria infection or received antimalarial drugs from the community system, which both can lead to false-positive pigment and rapid test results.

With automated malaria detection, it is not possible to differentiate between treatment failure, re-infection or simply pigment tests remaining positive after successful treatment. In high-risk malaria areas, with numerous potential infectious bites per week, the test can therefore not be used for diagnosing malaria. However, the high specificity serves as an accurate exclusion criterion for malaria during this season in the case pigment results are negative. In low-risk malaria areas, the test can be used for diagnosing malaria because of the high sensitivity. Because the malaria pigment test is more sensitive than microscopy and requires less time, it could decrease workload and patient waiting time.

Sensitivity and specificity of the test were found to be influenced by the cut-off point with respect to the number of pigment dots regarded as a positive test result. Our data suggest that a cut-off point of ≥1 dot(s) results in the most discriminating power.

To our knowledge, this is the first study reporting the inconsistency of malaria pigment results. When a sample was run twice, it resulted in a change of diagnosis in 14%. This is an important limitation of the test and a potential target for improving the accuracy.

Malaria pigment was negatively correlated with haemoglobin and haemoglobin levels have to be closely observed in patients with a high dot count (>14 dots). As malaria pigment results from the breakdown of haemoglobin by malaria parasites, it can be hypothesized that when a large amount of haemoglobin complexes are broken down, the haemoglobin level will fall and the haemozoin level will rise, resulting in an increase of pigment dots. Unfortunately, the pigment dot count cannot be used as a marker for severity of malaria infection, because there was no correlation between the pigment dot count and the parasitaemia as reported earlier (Sullivan & Meshnickb 1996; Dromigny et al. 2005).

We conclude that automated detection of malaria pigment is a useful malaria diagnostic tool in this semi-rural area. In low-risk malaria season, the test can be used for diagnosing malaria because of the high sensitivity. In high-risk malaria season, the test can be used for excluding malaria in case of a negative pigment result because of the high specificity. In our setting, the running costs of both tests were in favour of the malaria pigment test and using this test in low-risk malaria season would decrease hospital costs, potentially resulting in a cost reduction of 32%. The inconsistency of the dot count with repetition of the test and the long-time pigment results remain positive limit its usefulness.

Acknowledgements

We would like to thank the management team of Onandjokwe Lutheran Hospital and the management board of the National Institute of Pathology for approval of the study, Sam Davids for providing us with the 2002 data and all NIP staff for assisting us with collecting the data. This study was financed by ‘Stimuleringsfonds’, Dutch Society of Tropical Medicine. AJ de Langen was sponsored by ‘Dittmerfonds’, Free University, Amsterdam.

Ancillary