506

TW.NSACXIONS or THE ROYAL SCZIETY OF TROPICAL MEDICINE AND HYGIENE (1991) 85, 506-510

A quantitative approach to the relationship between Wuchereria bancrofti microfilaria counts by venous blood filtration and finger-prick blood films*

Mohamed Sabty United States Naval Medical Research Unit No. 3, Cairo, Egypt

Abstract Counts of nocturnally periodic Wuchereria bancrofti

microlilariae (ml) in 20 mm3 finger-prick blood films were comnared with membrane counts after filtration of 1 ml of venous blood for their efficacy in determin- ine the nrevalence of Bancroftian filariasis. The techniqu; used for spreading and staining the blood fihns was critically important to the sensitivity of this screening procedure. There was good agreement between the 2 methods. Four statistical approaches were used to measure the correlation between the 2 sampling methods: 2 linear regression methods using untransformed and transformed data, and 2 non- parametric rank correlation methods. Based on the statistical analyses, this study strongly supports the general observation that finger-prick blood samples tend to contain more mf than equivalent volumes of venous blood, at mf densities high enough to be detectable by the linger-prick technique. It also demonstrates that tiger-prick samples provide good estimates of mf densities and prevalence of infection except in areas of very low mf densities, where the membrane filtration method would provide a more accurate estimate of prevalence. Regression analyses using untransformed and transformed data, and the rank-correlation tests, demonstrated a strong statisti- cally significant correlation (P<O*OOl) between venous and finger-prick mf counts.

Introduction Different volumes of linger-prick blood have been

used in survevs to determine the orevalence of Bancroftian filariasis, ranging from 26 to 100 mm3 (ABARU & DENHAM, 1976; SASA, 1967; DENHAM et al., 1971). Similarly, the spreading technique used to prepare blood films has yet to be standardized. Researchers have annlied the blood in a thick circle or square of 3 cm2, or applied it in a single strip or 3 parallel strips of 20 mm3 each (SASA, 1967). Each of these methods necessitates dehaemoglobinization of red cells by dipping the blood fihn in water and it has been calculated by DENHAM et al. (1971) and WEBBER (1977) that about 30-40% of the microfilariae (mf) in fihns were lost during this nrocedure. SABRY (1987) minimized mf loss bi spreading the 20 mm3 ‘blood sample into a rectangular shape, 1.3x3.8 cm, and staining without prior dehaemoglobinization or rins-

*This work was supported by the Naval Medical Research and Development Command, Naval Medical Command, National Capital Region, Bethesda, Maryland, USA, work unit no. 3M162770A870.AR.322. The opinions and asser- tions contained herein are the private ones of the authors and are not to be construed as official or as reflecting the views of the Navy Department, Department of Defense, or the US Government.

Correspondence and reprint requests should be addres- sed to : Research Publications Branch, US Naval Medical Research Unit No. 3., FPO New York 09527-1600, USA.

ing thereafter. LAGRAULET et al. (1972) indicated that the highest mf concentrations were obtained in finger-prick samples although results from ear-prick blood suggest that as many as 9.5% of the negative finger-prick samples may actually have been positive.

ABARU & DENHAM (1976) used the lysed blood smear method and estimated the overall efficiency to be 100% (wet smear), 97.1% (stained smear), and 97.7% (counting chamber). The counting chamber technique revealed approximately 2.4 times as many mf as thick blood films (MCMAHON et al., 1979; CRANS, 1972; DENHAM et al., 1971; DESOWITZ & SOUTHGATE, 1973; SOUTHGATE, 1973, 1974).

PARTONO et al. (1973) reported that the concentra- tion method for detecting mf was as efficient as the membrane filtration technique. In this procedure blood is haemolysed by 2% saponin or 10% TeepoP (BELL, 1967) or laked with 2% formalin (KNOTT, 1939), and then centrifuged before microscopical examination.

The membrane filtration technique, using 1 ml of venous blood, is very accurate, mf loss is nii; and it is capable of detecting low densities of mf below the level of sensitivity of the finger-prick method. SOUTH- GATE (1974) found that onlv 38.8% of his survev population &as mf positive bv the counting chamber _ _ technique, compared to 67.8% when the membrane filtration technique was used. The nrimarv drawback of the filtration technique is that it is difficult to use in survey studies and requires more time and effort.

The present study attempted to measure the relationships between venous blood filtration and finger-prick blood films for measurement of Wuchereria bancrofti microfilaraemia.

Materials and Methods A household by household survey was made in the

Marsafa area of Qalubyia governorate, Egypt, an endemic area for filariasis, to identify microlilaraemic individuals from both sexes who were 6 years of age or older. After initial screening 21 volunteers with varying ranges of microfilaraemia were chosen ran- domly to participate in the study. Informed consent was signed by each volunteer before giving his/her blood sample.

A 20 mm3 finger-prick blood sample, and a 1 ml venous blood sample drawn into a tube containing a few crvstals of ethvlenediaminetetraacetic acid were taken simultaneously from each volunteer. All blood collections were carried out at 2100-2400 h to coin- cide with the nocturnal periodicity of microfilaraemia.

The 20 mm3 sample was spread on a slide in a rectangle of 103~3.8 cm, allowed to dry for 36 h, and stained for 40 min in a 1:20 solution of Giemsa’s stain, without prior dehaemoglobinization or rinsing after staining (SABRY, 1987).

The 1 ml of venous blood was mixed with the lysing solution used in counting white blood cells,

507

filtered through a Nucleporem membrane filter (25 mm diameter, 5 pm pore size), and stained for 10 min in a 1:20 solution of Giemsa’s stain (SABRY, 1987).

The mf counts from the blood film and membrane filter of each volunteer were assigned to one of 7 arbitrarily designated density classes: class I, l- 10 mf/ml venous blood; class II, 11-50; class III, 51-100; class IV, 101-500; class V, 501-1000; class VI, 1001-5000; class VII, >5000.

Because the coefficient of correlation (r) measures the strength of a relation between 2 variables but not the agreement between them, and a high correlation may conceal considerable lack of agreement between two methods or techniques, I found it more appropri- ate to measure also the agreement between venous and finger-prick blood sampling methods by estimating the limits of agreement as recommended by ALTMAN & BLAND (1983).

Results Table 1 summarizes the characteristics of the 21

volunteers classified into the 7 density classes. Table 2 presents data on observed mf counts per 20 mm3 finger-prick blood film and the corresponding counts expected in an equivalent volume of venous blood, obtained by dividing the count resulting from filtra- tion of 1 ml venous blood by 50.

Results of statistical analyses of these data are shown in Tables 3 to 5. Because both data sets were highly skewed, analyses have been performed by 4 different methods: 2 linear regression techniques and 2 non-parametric rank correlation tests. Table 3

Table 1. Summary characteristics of the 21 human volunteers participating in the comparative evaluation of venous blood filtration and finger-prick blood films for estimation of microfilaraemia levels in Bancroftian filariasis

Microfilarial count 20 mm3

Age Density One ml No. (years) Sex” class venous blood

tinge;epick

M :; :t F

I

13 13 F II 5

12 :z iz

:; 12 M III

16 :i iFI

:t 17

iti IV

11 :: M 21 25 F V

: :: ii

; :: M VI

: 24 E

14 :: M” 3 21 M VII 1 13 M

4 7

12 13 19

68 96 98

114 126 136 609 642 726

1364 1541 1987 2373 3894 5777 7343

8

192 119

shows simple linear regression analysis of the 2 sets of data, including the zero values obtained from 3 of the 21 finger-prick blood fihns and regression analysis using the log,, (n+l) transformation of data in an attempt to normalize the data and allow for log- transformation of zero values before analysis. Simple inspection of the data grouped by density class in Table 2 revealed an apparently perfect rank correla- tion between class and expected counts in venous blood; thus an analysis was not performed. However,

Table 2. Relationship between observed microtiiarial counts p&r 20 mm3 Iinger-prick blood and expected counts per 20 mm3 of venous blood (a=21)

No.

19 18

Density class

I

Observed count Expected count x Mean Y Mean

0 0.08 0 0.14 0.1

13 5

12

17 20 16

10 15 11

21 2 4

8 7 9 6

14

3 1

II 0 2b lb

III i%

IV 8b 13b

3b

V 7 20b

VI 24 b

is 81b

141b

VII 192b 119

0.24 0.26

1.0 0.38 0.3

1.36 1.92

3.7 1.96 1.7

2.28 2.52

8.0 2.72 2.5

12.18 12.84

15.7 14.52 13.2

27.28 30.82 39.74 4746

68.0 77.88 44.6

115.54 155.5 146.86 131.2

‘Microfilarial counts in 20 mm3 of venous blood were obtained by dividing the observed counts in 1 ml of blood (Table 1) by 50.

bObserved count >expected count.

Table 3. Analysis of the relationship behveen observed microtila- rial couats in 20 mm’ of finger-prick blood (x) sad calculated 20 mm3 counts derived from filtration of venous blood (y), a=21

Arithmetic data Regression equation T

y=1.35+0.69x 0.92

Standard error (b) 95% confidence limits (b) P Arithmetic mean: ir, 9 Limit of x as y-0 Limit of y as x-0

Logarithmic transformation of data Regression equation log,, (y+l)= T

‘(19)

10.13 0.069 0.55-0.84

<o+lO1 35.05, 25.67

-1.95 1.35

Standard error (b) 95% confidence limits (b) P Arithmetic mean: log (?+l),

log (Y+l)

-0.09+0.95 log,&+ 1) 0.96

15.35 0.06 0.82-1.08

<OQOl 1.56. 1.43

Limit of x as y-0 0.25 Limit of y as x-*0 -0.19 “F=female, M=male.

508

Table 4 shows the results of Spearman’s and Ken- dall’s rank correlation tests using all the 21 individual results given in Table 2. As expected, a strong correlation between the 2 assays was found, using these two non-parametric tests. Table 5 presents both transformed and untransformed linear regression analyses of the grouped density class data set out in Table 2.

To assist in interpreting these Tables, scatter diagrams of the 21 pairs of data points (observed finger-prick versus calculated venous mf counts) are shown in Fig. 1 (untransformed data) and Fig. 2 (logarithmic transformed data).

Table 4. Analysis of the relationshiq between observed microfilarial counts in 20 mm of finger- prick blood and calculated 20 mm3 counts derived from filtration of venous blood, using non-paramet- ric statistical methods, n=21

t (Y)

200

160

120

00.

40.

6 I

Factor Result

Spiarman’s rank correlation test 0.97

P <O*OOl 0.98 19.56

<O~OOl Kydall’s rank correlation test

: SND P

194 0.96

<O.OOl 6.07

<O~OOOl

Table 5. Analysis of results presented in Table 2, using both arithmetic untransformed and log-transformed data of microfikial density classes, o=7

Parameters Values

Untransformed data Regression equation y=-2.28+0,83x r 0.995

t(s) 23.18 Standard error (b) 0.036 95% confidence limits (b) 0.74-0.92 P <0~001 Arithmetic mean ,, y 35.98, 27.67 Standard deviation (n-1) x, y 57.83, 48.36 Coefficient of variation x, y 16Og%,174g%

Log,, (x+1) (y+l) transformed data Regression equation log,,(y+l)=-o~15+0~99 log,, (x+1) r 0.98

'(5) 12.12 Standard error (b) 0.082 95% confidence limits (b) 0.78 -1.20 P <0401 Arithmetic mean ho(~+ I)> ~wm(~+ 1) 1.57, 1.46

Standard deviation (n-l) . ~oglo(~+l)~ logm(Y+l)

Coefficient of variation ~okT,o(~+l), (y+l)

0.79, 0.80

50.3%, 54.8%

40 00 120 160 260

(xl +

Fig. 1. Scatter diagram using arithmetic data from microfilarial counts in 20 mm3 of finger-prick (x) and calculated from venous blood filtration (y) from the 21 volunteers listed in Tables 1 and 2:- calculated regression line; ---- 45” gradient 1 line (x=y). Inset shows atea from 0 to 4 enlarged.

2.4

2.0 1

t 1.6.

+ 1.2-

2 0 - :: 0.8.

0.4.

I” .

’ 0.0 I 0.0 0.4 0.8 1.2 1.6 2.0 2.4 2.8

log,g(x+l) --)

Fig. 2. Scatter diagram of logarithmic transfotmations of microfilatial counts in 20 mm’ of finger-prick blood [log,, (x+ l)] and calculated from venous blood filtration [log,, (y+ l)] for the 21 volumeen listed in Tables 1 and 2: - calculated regression line; ---- 45” gradient 1 line [log,, (x+l)=log10 (y+l)].

Discussion Most authors concluded that finger-prick blood

generally contains more mf than an equivalent volume of venous blood, when venous blood mf densities rise above 15 mf per ml (SOUTHGATE, 1974, 1984; MCMAHON et al., 1979; SHIBUYA et al., 1980; EBERHARD et al., 1988); below this critical mf density, many 20 mm3 finger-prick samples will be negative. Calculations using both grouped finger- prick and venous blood counts of the 7 microfilarial density classes presented in Table 2 to measure the

degree of agreement between the 2 sampling techni- ques gave the following results: mean difference @)=8*31, standard deviation (s)=10*75, standard error of a (SEd)=4*064,95% confidence intervals for ?I= - l-64 to H-25, and limits of agreement= - 12.77 to 29.38. All differences lay within the limits of agreement a+1.96s, hence the 2 measurement methods may be used interchangeably. Moreover, the SE5 of 4.064, which is relatively smalI, is another indication of a good estimate. Despite this, the venous mf counts tended to give a lower reading by between -1.6 and 18-3 mf/20 mm3, which supports the conclusion by many authors referred to above, that finger-prick blood tends to have higher mf densities than venous blood at mf densities high enough for 20 mm3 films to be positive. Both untransformed (Table 3 and Fig. 1) and transformed (Table 3 and Fig. 2) regression analyses confirmed these observa- tions. The slopes of the 2 regression lines were markedlv different, due to effects of zero values in the untransfbrmed regression. Both lines lay below the madient 1 line. but the untransformed line in Fin. 1 crossed the g&client 1 line at a level of x=y=q*4, while the loglo (n+ 1) transformed line in Fig. 2 lay below the eradient 1 line for its entire lenti. In Table 3, x always exceeded y after transformason, while x exceeded y only above the equivalence point of 4.4 mf per 20 mm3 in the untransformed regression. Rank order correlation was perfect in Table 2 for grouped density classes and hence was not tested statistically. Rank order correlation between venous and finger-prick techniques shown in Table 4, using data from all 21 individuals, was correlated at a level that was highly statistically sign&ant (P<O*OOl).

The repression narameters shown in Table 5 for mf counts grouped &to 7 density classes differed signi- ficantly from zero, and the slope of the transformed line was steeper (b=0+9) than that of the untrans- formed line (b=0*83). In addition, coefficients of variation for x and y were much higher with the untransformed data than with transformed values. This study therefore strongly supported the general observation that finger-prick blood samples tend to contain more mf than equivalent amounts of venous blood, at mf densities high enough for 20 mm3 Iihns to be positive. This study also demonstrated that finger-prick samples would provide good estimates of mf densities and prevalence of infection, and that the 2 measurement methods can be used interchangeably, except in areas of very low mf densities. In this situation, the membrane titration method would be more appropriate, since its greater sensitivity would nrovide a more accurate estimate of nrevalence.

In Egypt, where autogenous Culex &kns (=C.p. m0lestu.s~ is the main vector of filariasis. low densitv mf carriers?< 15 mf/ml venous blood or k2 mf/20 t&3 finger-phck blood) do not play an important role in filariasis transmission (SABRY, 1987). Thus this factor, together with the rel&tance of many Egyptians to aant nermission for venenuncture and the cost and Grne iequired to process venous samples, makes linner-brick blood films the method of choice for local iilaiiasis surveys. Although such surveys would miss a small number of low density mf carriers, they would be of minor epidemiological importance in the trans- mission of Bancroftian filariasis in Egypt (SABRY, 1987), and not in need of chemotherapy

References Abaru, D. E. & Denham, D. A. (1976). A comparison of the

efficiencv of the Nucleoore and Milliuore filtration systems. for detecting hcrofilariae. S&~theust Asiun Journal of Tropical Medicine and Public Health, 7, 367-369.

Altman, D. G. & Bland, J. M. (1983). Measurement in medicine: the analysis of method comparison studies. Statistician, 32, 307-317.

Bell, D. (1967). Membrane filters and microfilariae: a new technique. Annals of Tropical Medicine and Parasitology, 61, 22lL223.

Crans, W. J. (1972). A rapid technique for the determination of mrcrofiaraetnia infiluriuti carriers. Geneva: World Health Organization, mimeographed document WHO/FIL/721 98.

Denham, D. A., Dennis, D. T., Ponnudurai, T., Nelson, G. S. & Guy, F. (1971). Comparison of a counting chamber and thick smear methods of counting micro- filariae. Transactions of the Royal Society of Tropical Medicine and Hygiene, 65, 521-526.

Desowitz, R. S. & Southgate, B. A. (1973). Studies on filariasis in the Pacific: 2. The persistence of microfilar- aemia in diethylcarbamazine-treated populations of Fiji and western Samoa: diagnostic application of the mem- brane filtration technique. Southeast Asian 3ourttal of Tropical Medicine and Public Health, 4, 179-183.

Eberhard,, M. L., Roberts, I. M., Lammie, P. J. and Lowne, R. C., Jr (1988). Comparative densities of Wuchereria buncrofti microfilaria in paired samples of capillary and venous blood. Tropical Medicine and Parasiwlogy, 39, 295-298.

Knott, J. (1939). A method for making microfilarial surveys on day blood. Trans&ons of the Royal Society of Tropical Medicine and Hygiene, 33, 191-196.

Lagraulet, J., Pichon, G., Outin-Fabre, D., Stanghellini, A. & Moreau, J. P. (1972). Enquete epidkmiologique sur la filariose lymphatique aux Marquisas. Bulletin de la Sock% de Pathologic Exotique, 65, 447-455.

McMahon, J. E., Marshall, T. F. de C., Vaughan, J. P. & Abaru, D. E. (1979). Bancroftian filariasis: a comparison of micro6lariae counting techniques using counting chamber, standard slide. and me?brane (Nuc!epore) y3tra;;;Agnals of Troprcal Meduxte and Paranwlogy,

PartoAo, F., C&s, J. H., Purnomo & Oemijati, Sri (1973). Evaluation of thick smear, Knott and membrane filtra- tion methods for demonstrating microfilariae in blood. Tropical and Geographical Medicine, 25, 286-289.

Sabry, M. (1987). A study of the quantitative relationship between the blood densitv of Wuchereria bancrofti micro- fikzriae and developtnentbf infective larvae in Culex pipiens molestus in Ettvpt. Ph.D. thesis. Facultv of Medicine. University of liondon. ’ -

Sass, M. (1967). Microfilaria survey methods and analysis of survey data in lilariasis control programmes. Bulletin of the World Health Organization, 37, 629-650.

Shibuya, T., Cabrera, B. D., Tanaka, H., Valeza, F. S. & Instrella, R. (1980). Comparison of the blood film, Millipore filter and Nuclepore filter techniques for the detection of microfilaraemia in a field survey in the Philippines. .7apanese .‘fournal of Expetimental Medicine, 50, m-468: - - - -

Southaate, B. A. (1973). Studies on filariasis in the Pacific. I. A %elld trial of a counting chamber technique for the determination of microlilarial rates and densities. South- east AsianJouwtal of Tropical Medicine and Public Health, 4, 172-178.

Southgate, B. A. (1974). A quantitative approach to uarasitoloeical techniaues in Bancroftian filariasis and its effect on epidemiolog&l understanding. Transactions of the Royal Socieg of Tropical Medicine and Hygiene, 68, 177-185.

Southgate, B. A. (1984). Recent advances in the epidemiolo- gy and control of filarial infections including entomolo-

510

gical aspects of transmission. Ykansactrons of the Rqal estimation. Transactions of the Royal Society of Tropical Society of Tropical Medicine and Hygiene, 78, supplement, Medicine and Hygiene, 70, 537-538. 19-28.

Webber, R. H. (1977 [dated 19761). Comparison of counting chamber and measured blood films in microfilaria

Received 2 January 1990; revised 2 January 1991; accepted for publication 10 January 1991

Announcements 1

The Tropical Health and Education Trust Fellows of the Society have always been actively involved in many tropical countries in establishing

and developing medical schools and other training institutions. But some of these school:, particularly in poorer African countries, face severe hardships. Students have no books, there IS no foreign exchange for journals, equipment lacks spares, research cannot be supported and external aid is directed towards primary health care.

The Tropical Health Education Trust has started to relieve, with support from many individuals, Trusts and organizations, some of these disadvantages.

Basic books have been sent to all the rural hospitals in two African countries, sets of books have been given for students in a number of others. Links between medical schools overseas and home departments have been started with fellowships for students in training and research methods a!so.

The Tropical Health and Education Trust aims to extend support like this to more counmes, hospitals, medical schools and students and needs funds to do it: Fellows of the Society who would like to take this opportunity to help our colleagues overcome some of their obstacles can do so through a single gift, a four-year or a deposited covenant, or even through a legacy.

Trustees include: R. M. Anderson, K. P. W. J. McAdam, E. H. 0. Parry (Chairman), D. A. Warrell .

For more information about THET please write or telephone: 21 Edenhurst Avenue, Fulham, London, SW6 3PD, UK. Tel: 071 927 2411, Fax: 071 637 4314.