186 TRANSACTIONS OF THE ROYAL SOCIETY OF TROPICAL MEDICINE
Comparison density
of two
simple
methods
B. M. Greenwood and J. R. M. Armstrong
AND HYGIENE (1991) 85, 186-188
for determining
malaria
Medical Research Council Laboratories,
parasite
Fajara,
Banjul,
The
Gambia
Abstract Two simple methods for determining malaria parasite density which require only the examination of a thick blood film were compared with a more accurate method which involves determination of the red blood cell count and measurement of the level of infection in red blood cells by examination of a thin blood film. In one method, the number of Darasites present per white blood celiis counted and t&s figure multiulied bv 8000 (an average white blood cell count per $) to give th; parasit; density. In the other method, the number of parasites present per high power microscope field is determined and the parasite density calculated from this value and the assumed volume of blood present in one high power field. The latter method proved to be more accurate than that based on deteimination of the parasite/white blood cell ratio, probably because the variability in the volume of blood used in preparation of thick blood films was less than the variabilitv of the white blood cell count in the population s&died.
Introduction The relationship between the clinical severity of malaria and the density of malaria parasitaemia is not straightforward, perhaps due in part to the sequestration of parasites during Plasmodium falciparum infections. However, high parasitaemia is, in general, more likely to be a cause of illness than low parasitaemia. For this reason, determination of malaria parasite density is useful both in the assessment of an individual subject with symptoms and malaria parasitaemia and in epidemiological surveys, the objectives of which include determination of the likely clinical impact of malaria on a study community. Accurate determination of parasite density requires a red blood cell (RBC) count and accurate-measurement of the level of RBC infection as assessed bv examination of a thin blood film. It is usualli recommended that at least 100 000 RBC should be counted to give a true value. Thus, this method is time-con&&g and it is not suitable for large-scale field studies. Two indirect methods are used more widely in busy clinics and during field surveys. In the first, the number of parasites per white blood cell (WBC) in a thick blood film is recorded and parasite density is determined by multiplying this figure by 8000, an average WBC count per ~1 (TRAPE, 1985). In the second method, the average number of para$ites per high Power microscone field (HPF) of a thick blood -film is measured -and the‘ par&e density calculated from this value and the estimated amount Address for correspondence: Dr B. M. Greenwood, Medical Research Council Laboratories, P.O. Box 273, Banjul, The Gambia.
of blood present in a thick blood film (MOLINEAUX & GRAMICCIA. 19801. Both these indirect methods are subject to error. ?he first method will be inaccurate when the WBC count differs markedly from the assumed value of 8000 per ~1, as may occur in children with malaria, and the second method will be inaccurate when thick blood smears are prepared incorrectlv. To determine which method is the least inaccuratd we measured parasite density by both methods in 50 Gambian subiects attending an outpatient clinic and compared the results obtayned with those of a more accurate test which employs measurement of the RBC count and careful examination of a thin blood film.
Subiects and Methods Subjects Fifty patients with malaria parasitaemia who presented at the out-natient denartment of the Medical Research Council Laboratories, Fajara, The Gambia, during the malaria transmission season were studied. Subjects were selected bv preliminarv screening to cover a wide range of paraiitaemias. -Some pati&ts had severe clinical malaria whilst others had onlv minor symptoms which were unlikely to be related tb their parasitaemia. All patients had a P. falciparum infection. Measurement of malaria parasite density Method 1. The RBC count was determined using a Coulter ZF counter. A thin blood film stained with Giemsa’s stain was examined and the level of infection in 100 000 RBC recorded. All thick and thin films were examined by one observer. The parasite rate was determined bv multinlvine the RBC count bv thepercentage of ‘RBC iniected’ with an asexual parasite. The value obtained in this way was considered to be the ‘true’ parasite density. Method 2. One hundred HPF of the Giemsa-stained thick blood film were examined using a Leitz Laborlux@ microscoDe with X 10 eveDieces and x 100 objective. The blood film was examined systematicalIv to ensure that each section of the film was covered: the film was scanned from side to side and then fro& top to bottom until 100 HPF had been viewed. The number of parasites and WBC present in 100 HPF was determined and the average number of parasites per WBC calculated. This figure was multiplied by 8000 (the assumed WBC count) to give the parasite density (method 2a). The accuracy of this method was also investigated when a ‘true’, rather than an assumed, WBC count determined using a Coulter counter was used (method 2b). Method 3. One hundred HPF of a Giemsa-stained thick film were examined as described above and the number of HPF containing one or more parasites per HPF determined. When each field contained one or
187 more parasites, l&SO HPF were examined to determine the number of parasites per HPF (the number of fields counted was determined by the parasite density-only 10 were counted when this was very high). Preliminary experiments carried out using accurately measured volumes of blood showed that the volume required to give a satisfactory thick blood film was in the range of 5-8 4 and that the volume of blood present in one HPF of a well-prepared thick film was about 0402 ~1. Thus, the average number of parasites per HPF was multiplied by 500 to give an estimated parasite density.
FreqUeiWy 25
20
Statistical methods
To compare tests, a simple approach using graphical methods and analysis of variance (ANOVA) was adopted, based on that proposed by ALTMAN & BLAND (1983). The bias and error for each method were assessed,and ways of increasing accuracy and precision using red and white blood cell counts were investigated. Due to the skewed nature of the parasite and cell count data, logarithmic transformations were used throughout. Results Classification of parasite density
0.0
0.5
1.0 True
count
1.0 True
count
1.5
2.0 I Estimated
2.5 count
1.5
2.0 / Estimated
2.5 count
Frequency 25,
20
15
10
A simple assessmentof the ability of methods 2a, 2b and 3 to measure parasite densities accurately is shown in the Table, in which true parasite densities were arranged in 3 broad bands. Method 3 was more successful at identifying high and low levels of parasitaemia than method 2a; the rate of misclassification for medium level parasitaemiaswas the samefor both methods. For high level parasitaemias, 50 000 per ~1or more, methods 2a and 3 had a specificity of 100%. Method 3 had a sensitivity of 93% whilst that of method 2a was only 78%. Method 2b, in which the ‘true’ rather than an assumedWBC count was used, had a specificity of 96% and a sensitivity of 89%.
5
01
1 0.0
0.5
Frequency
251
15
Table. Assessment of the accuracy of two methods of determining malaria parasite density from examination of a thick blood film 10
True count
WOOO
Total 5
Method 2a” 4000 5-50000 >5c@OO Total
4 2 0 6
MethodZb’ i5000 5-5OOQO >50000 Total Method 3b 50000 Total
1 16 2;
0 0 21 21
5 I8 27 50
5 1 0 6
0 16 3 19
0 1 24 25
5 18 27 50
5 2 0 7
0 16 2 18
0 0 25 25
5 18 27 50
‘Methods 2a and 2b involved determination of the parasite/white blood cell ratio, using an assumed count of 80001pl in method 2a and an actual count in method 2b; parasite counts are expressed per pl (see text for further details). bMethod 3 involved determination of the number of parasites per high power microscope field; parasite counts are expressed per ~1 (see text for further details).
200 ML
-0.0
0.5
Figure. The error of two methods of measuring malaria parasite density by examination of a thick blood film as shown by the frequency distribution of true counts/estimated counts. Logarithmic scales are used. Upper: method 2a based on determination of the parasite/white blood cell ratio assuming a white blood cell count of 8OOO/pl. Middle: method 2b based on determination of the parasite/ white blood cell ratio, using a measured white blood cell count. Lower: method 3 based on determination of the number of parasites present per high power microscope field.
Bias and error
The distribution of errors for the 3 methods (the frequency distribution of the ratio of true to estimated counts) is shown in the Figure. Methods 2a and 2b are both subject to larger error than method 3. Bias was assessedby calculating the mean difference in count (true minus estimated, after log transformation of the
188 data). All 3 methods tended to underestimate the parasite density; method 2a gave on average 87% of the true density, method 2b gave 96% and method 3, 86%. This bias was sienificant for method 3 but not for methods 2a or 2b, &e more to the smaller errors for method 3 than to a greater degree of bias (paired c tests: 49 degrees of freedom, t=351, P
Comparison density
of two
simple
methods
B. M. Greenwood and J. R. M. Armstrong
AND HYGIENE (1991) 85, 186-188
for determining
malaria
Medical Research Council Laboratories,
parasite
Fajara,
Banjul,
The
Gambia
Abstract Two simple methods for determining malaria parasite density which require only the examination of a thick blood film were compared with a more accurate method which involves determination of the red blood cell count and measurement of the level of infection in red blood cells by examination of a thin blood film. In one method, the number of Darasites present per white blood celiis counted and t&s figure multiulied bv 8000 (an average white blood cell count per $) to give th; parasit; density. In the other method, the number of parasites present per high power microscope field is determined and the parasite density calculated from this value and the assumed volume of blood present in one high power field. The latter method proved to be more accurate than that based on deteimination of the parasite/white blood cell ratio, probably because the variability in the volume of blood used in preparation of thick blood films was less than the variabilitv of the white blood cell count in the population s&died.
Introduction The relationship between the clinical severity of malaria and the density of malaria parasitaemia is not straightforward, perhaps due in part to the sequestration of parasites during Plasmodium falciparum infections. However, high parasitaemia is, in general, more likely to be a cause of illness than low parasitaemia. For this reason, determination of malaria parasite density is useful both in the assessment of an individual subject with symptoms and malaria parasitaemia and in epidemiological surveys, the objectives of which include determination of the likely clinical impact of malaria on a study community. Accurate determination of parasite density requires a red blood cell (RBC) count and accurate-measurement of the level of RBC infection as assessed bv examination of a thin blood film. It is usualli recommended that at least 100 000 RBC should be counted to give a true value. Thus, this method is time-con&&g and it is not suitable for large-scale field studies. Two indirect methods are used more widely in busy clinics and during field surveys. In the first, the number of parasites per white blood cell (WBC) in a thick blood film is recorded and parasite density is determined by multiplying this figure by 8000, an average WBC count per ~1 (TRAPE, 1985). In the second method, the average number of para$ites per high Power microscone field (HPF) of a thick blood -film is measured -and the‘ par&e density calculated from this value and the estimated amount Address for correspondence: Dr B. M. Greenwood, Medical Research Council Laboratories, P.O. Box 273, Banjul, The Gambia.
of blood present in a thick blood film (MOLINEAUX & GRAMICCIA. 19801. Both these indirect methods are subject to error. ?he first method will be inaccurate when the WBC count differs markedly from the assumed value of 8000 per ~1, as may occur in children with malaria, and the second method will be inaccurate when thick blood smears are prepared incorrectlv. To determine which method is the least inaccuratd we measured parasite density by both methods in 50 Gambian subiects attending an outpatient clinic and compared the results obtayned with those of a more accurate test which employs measurement of the RBC count and careful examination of a thin blood film.
Subiects and Methods Subjects Fifty patients with malaria parasitaemia who presented at the out-natient denartment of the Medical Research Council Laboratories, Fajara, The Gambia, during the malaria transmission season were studied. Subjects were selected bv preliminarv screening to cover a wide range of paraiitaemias. -Some pati&ts had severe clinical malaria whilst others had onlv minor symptoms which were unlikely to be related tb their parasitaemia. All patients had a P. falciparum infection. Measurement of malaria parasite density Method 1. The RBC count was determined using a Coulter ZF counter. A thin blood film stained with Giemsa’s stain was examined and the level of infection in 100 000 RBC recorded. All thick and thin films were examined by one observer. The parasite rate was determined bv multinlvine the RBC count bv thepercentage of ‘RBC iniected’ with an asexual parasite. The value obtained in this way was considered to be the ‘true’ parasite density. Method 2. One hundred HPF of the Giemsa-stained thick blood film were examined using a Leitz Laborlux@ microscoDe with X 10 eveDieces and x 100 objective. The blood film was examined systematicalIv to ensure that each section of the film was covered: the film was scanned from side to side and then fro& top to bottom until 100 HPF had been viewed. The number of parasites and WBC present in 100 HPF was determined and the average number of parasites per WBC calculated. This figure was multiplied by 8000 (the assumed WBC count) to give the parasite density (method 2a). The accuracy of this method was also investigated when a ‘true’, rather than an assumed, WBC count determined using a Coulter counter was used (method 2b). Method 3. One hundred HPF of a Giemsa-stained thick film were examined as described above and the number of HPF containing one or more parasites per HPF determined. When each field contained one or
187 more parasites, l&SO HPF were examined to determine the number of parasites per HPF (the number of fields counted was determined by the parasite density-only 10 were counted when this was very high). Preliminary experiments carried out using accurately measured volumes of blood showed that the volume required to give a satisfactory thick blood film was in the range of 5-8 4 and that the volume of blood present in one HPF of a well-prepared thick film was about 0402 ~1. Thus, the average number of parasites per HPF was multiplied by 500 to give an estimated parasite density.
FreqUeiWy 25
20
Statistical methods
To compare tests, a simple approach using graphical methods and analysis of variance (ANOVA) was adopted, based on that proposed by ALTMAN & BLAND (1983). The bias and error for each method were assessed,and ways of increasing accuracy and precision using red and white blood cell counts were investigated. Due to the skewed nature of the parasite and cell count data, logarithmic transformations were used throughout. Results Classification of parasite density
0.0
0.5
1.0 True
count
1.0 True
count
1.5
2.0 I Estimated
2.5 count
1.5
2.0 / Estimated
2.5 count
Frequency 25,
20
15
10
A simple assessmentof the ability of methods 2a, 2b and 3 to measure parasite densities accurately is shown in the Table, in which true parasite densities were arranged in 3 broad bands. Method 3 was more successful at identifying high and low levels of parasitaemia than method 2a; the rate of misclassification for medium level parasitaemiaswas the samefor both methods. For high level parasitaemias, 50 000 per ~1or more, methods 2a and 3 had a specificity of 100%. Method 3 had a sensitivity of 93% whilst that of method 2a was only 78%. Method 2b, in which the ‘true’ rather than an assumedWBC count was used, had a specificity of 96% and a sensitivity of 89%.
5
01
1 0.0
0.5
Frequency
251
15
Table. Assessment of the accuracy of two methods of determining malaria parasite density from examination of a thick blood film 10
True count
WOOO
Total 5
Method 2a” 4000 5-50000 >5c@OO Total
4 2 0 6
MethodZb’ i5000 5-5OOQO >50000 Total Method 3b 50000 Total
1 16 2;
0 0 21 21
5 I8 27 50
5 1 0 6
0 16 3 19
0 1 24 25
5 18 27 50
5 2 0 7
0 16 2 18
0 0 25 25
5 18 27 50
‘Methods 2a and 2b involved determination of the parasite/white blood cell ratio, using an assumed count of 80001pl in method 2a and an actual count in method 2b; parasite counts are expressed per pl (see text for further details). bMethod 3 involved determination of the number of parasites per high power microscope field; parasite counts are expressed per ~1 (see text for further details).
200 ML
-0.0
0.5
Figure. The error of two methods of measuring malaria parasite density by examination of a thick blood film as shown by the frequency distribution of true counts/estimated counts. Logarithmic scales are used. Upper: method 2a based on determination of the parasite/white blood cell ratio assuming a white blood cell count of 8OOO/pl. Middle: method 2b based on determination of the parasite/ white blood cell ratio, using a measured white blood cell count. Lower: method 3 based on determination of the number of parasites present per high power microscope field.
Bias and error
The distribution of errors for the 3 methods (the frequency distribution of the ratio of true to estimated counts) is shown in the Figure. Methods 2a and 2b are both subject to larger error than method 3. Bias was assessedby calculating the mean difference in count (true minus estimated, after log transformation of the
188 data). All 3 methods tended to underestimate the parasite density; method 2a gave on average 87% of the true density, method 2b gave 96% and method 3, 86%. This bias was sienificant for method 3 but not for methods 2a or 2b, &e more to the smaller errors for method 3 than to a greater degree of bias (paired c tests: 49 degrees of freedom, t=351, P
