What does a single determination of malaria parasite density mean? A longitudinal survey in Mali

Tropical Medicine and International Health

volume 5 no 6 pp 404–412 june 2000

© 2000 Blackwell Science Ltd404

What does a single determination of malaria parasite densitymean? A longitudinal survey in Mali

Véronique Delley1, Paul Bouvier1, Norman Breslow1,3, Ogobara Doumbo2, Issaka Sagara2, Mahamadou Diakite2,Anne Mauris1,Amagana Dolo2 and André Rougemont1

1 Institute of Social and Preventive Medicine, University of Geneva, Switzerland

2 Faculty of Medicine, Pharmacy and Dentistry, University of Bamako, Mali

3 School of Public Health and Community Medicine, University of Washington, Seattle, USA

Summary Temporal variations of blood parasite density were evaluated in a longitudinal study of young, asympto-

matic men in a village with endemic malaria in Mali (West Africa). Our main intention was to challenge the

value of a single measure of parasite density for the diagnosis of malaria, and to define the level of endem-

icity in any given area. Parasitaemia and body temperature were recorded three times a day in the wet season

(in 39 subjects on 12 days) and in the dry season (in 41 subjects on 13 days). Two thousand nine hundred and

fifty seven blood smears (98.5% of the expected number) were examined for malaria parasites. We often

found 100-fold or greater variations in parasite density within a 6-hour period during individual follow-up.

All infected subjects had frequent negative smears. Although fever was most likely to occur in subjects with a

maximum parasite density exceeding 10000 parasites/mm3 (P 5 0.009), there was no clear relationship

between the timing of these two events. Examples of individual profiles for parasite density and fever are

presented. These variations (probably due to a ‘sequestration-release’ mechanism, which remains to be

elucidated) lead us to expect a substantial impact on measurements of endemicity when only a single sample

is taken. In this study, the percentage of infected individuals varied between 28.9% and 57.9% during the dry

season and between 27.5% and 70.7% during the wet season. The highest rates were observed at midday, and

there were significant differences between days. Thus, high parasite density sometimes associated with fever

can no longer be considered as the gold standard in the diagnosis of malaria. Other approaches, such as

decision-making processes involving clinical, biological and ecological variables must be developed, es-

pecially in highly endemic areas where Plasmodium infection is the rule rather than the exception and the

possible causes of fever are numerous.

keywords diagnosis, malaria, Mali, parasitaemia, Plasmodium

correspondence Véronique Delley, Institute of Social and Preventive Medicine, Centre Médical

Universitaire, 1 rue Michel-Servet, 1211 Genève 4, Switzerland. E-mail: [email protected]

Introduction

The detection of parasites in peripheral venous or capillary

blood has always been considered an indispensable basis for

the definition and diagnosis of malaria. Nonetheless, the true

significance of parasitaemia remains poorly understood.

Whether in the study of thresholds (Trape et al. 1985; Genton

et al. 1994) for the estimation of fever risk (Smith et al.

1994a) or in the simple determination of parasite density,

data collection has usually been limited to a single time

point. Little is known about the natural variations in parasite

density in the peripheral blood during the course of a day or

a week. Although it is likely that the potential for and severity

of a malaria attack depends on an individual’s total parasite

load, the parasitaemia measured in a single sample of periph-

eral blood at any given moment is unlikely to reflect this total

accurately. The capacity of haematocytes infected by

Plasmodium falciparum to agglutinate and to adhere to the

capillary endothelium leads to their sequestration in tissues

of the spleen, brain, placenta and other organs. At present,

the mechanisms responsible for sequestration and release of

infected haematocytes are largely unknown. It is unclear, for

example, whether the numbers of parasitized red cells in the

peripheral blood and the numbers in sequestration sites

TMIH566

Tropical Medicine and International Health volume 5 no 6 pp 404–412 june 2000

V. Delley et al. Single determination of malaria parasite density

remain at a constant ratio over time within an individual or

even between individuals.

The phenomenon of sequestration is probably the source

of both past and current reports of typical, severe malarial

attacks that occur with low (even zero) levels of peripheral

parasitaemia, or of high levels . 100 000 parasites/mm3 in

subjects who present with no symptoms at all (Hogh et al.

1993). Our ignorance of the distribution of parasitized

haematocytes between the central (tissues) and the peripheral

compartments (capillary and venous blood), and of the

mechanisms that determine this distribution, makes it im-

possible to determine the malarial origin of a fever episode

from the observation of parasitaemia alone and moreover at

a single occasion. Despite the recent accumulation of evi-

dence in support of this viewpoint (Bouvier et al. 1997;

Farnert et al. 1997) positive parasite counts from single

measurements continue to be used in practice to verify the

malarial origin of fever and to recommend treatment, at least

in the rare situations where a microscope is available.

During the 1950s, several longitudinal studies of adults

living in endemic zones (Miller 1958; Bruce-Chwatt 1963)

were conducted to understand better how parasitaemia varies

over time. More recently, a daily parasitological examination

in asymptomatic children has shown alternatively high and

low parasite densities. However, as blood samples were only

collected on a weekly or, at best, a daily basis (Gendrel et al.

1992; Farnert et al. 1997), the natural history of same-day

variations required further investigation. Our study sought to

evaluate the following hypotheses:

In a single individual, parasite density varies spontaneously

during the course of a several days follow-up; such variations

can lead to an erroneous estimation of the community load

of malaria infection; high parasite densities may be observed

in symptom-free individuals, while serious malarial attacks

may occur among those with no detectable parasites; and

there is no obligate temporal correlation between the occur-

rence of fever and peripheral parasite density.

The study was conducted from a strictly operational per-

spective. As previously, blood samples were examined follow-

ing the best possible clinical protocol (Rougemont et al. 1991;

Bouvier et al. 1997).

Materials and methods

Study area

The study was conducted in Bougoula, a rural village with a

population of 3200, situated 10 km east of Sikasso, the

capital of the third administrative region of the Mali

Republic. Most of the inhabitants belong to the Senoufo

ethnic group and have a traditional lifestyle mainly based on

agriculture. The pattern of malaria transmission is typical of

the savannah ecological zone. The area has a wet and a rainy

season, with an average annual rainfall of 1100 mm. Most of

this (75%) occurs between July and September. According to

a study conducted between June 1992 and September 1993 in

Pimperena, 15 km north of Sikasso, malaria transmission is

intense between July and November. Virtually no rainfall

occurred and no transmission was detectable from December

to June. The main vectors of malaria are mosquitoes from

the Anopheles gambiae sl complex (about 95% A. gambiae ss

and 5% A. arabiensis) (Toure et al. 1998; unpublished data).

The area had no active programme of spraying or insecticide-

impregnated bednets while the study was carried out. The

village was also chosen because of its very limited access to

health care.

Laboratory and field methods

Previous studies demonstrated the need to separate the data

according to dry and wet season, and to regard season itself

as a major risk factor (Greenwood et al. 1987; Bouvier et al.

1997). Thus we conducted our study in two phases. The

first took place during 12 days at the end of the dry, low-

transmission season (May 1994) and the second during

13 days at the end of the wet, high-transmission season

(October 1994). A religious festival caused temporary inter-

ruption of the study during the third day of the dry season

phase.

Young men without signs of disease or incapacity were

enrolled in the study. They were 20–35-year-old (mean

24 years). At the beginning of each phase, the volunteer sub-

jects completed a questionnaire with the assistance of a trans-

lator. The questionnaire covered basic personal data and the

immediate past history of fever episodes and treatment with

antimalarial drugs. Each day the subjects were examined at

0700, 1300 and 1900 h, at which time a finger-prick blood

sample was drawn and temperature was taken orally using a

standard electronic thermometer (Terumo®; Terumo

Corporation, Somerset, NJ, USA). Once each day, during the

morning examination, the packed cell volume (PCV or

haematocrit) was measured and the spleen was palpated and

classified according to Hackett (Gilles 1993). Any possible

metabolites of chloroquine in the urine were measured by the

Sakers–Solomon test (Mount et al. 1989) on the first and last

day of the study.

The protocol called for 3003 blood specimens to be taken

during the two phases, and 2957 (98.5%) were actually

obtained. Clinical and parasitological examinations were

done by the team of the Department of Epidemiology at the

Faculty of Medicine, Pharmacy and Odonto-Stomatology of

Bamako. Parasite density was determined by microscopic

examination of thick films using a 100 3 oil-immersion lens.

Estimation of the ratio of the number of parasites per

© 2000 Blackwell Science Ltd 405

Tropical Medicine and International Health volume 5 no 6 pp 404–412 june 2000

V. Delley et al. Single determination of malaria parasite density

leucocyte was based on counts of 300 cells in each thick

smear. The parasite density (parasites/mm3) was calculated by

assuming an average leucocyte concentration of 7500 leuco-

cytes/mm3. Thus, the recorded parasite density was obtained

by multiplying the observed number of parasites by 25. A

random sample of 10% of the slides was examined a second

time independently. A recent study of double-blind interpre-

tation of parasite slides in the laboratory of Bamako showed

excellent correlation between the two readings for negative

slides, good correlation for positive slides with medium and

high parasitaemia, but poor agreement for low parasite densi-

ties 5 100 (Doumbo OK, Diakité M, Dolo A, Diallo M &

Abdou M, unpublished data).

Fever was defined as a temperature $ 38 8C. When subjects

were found with fever, more detailed clinical examinations

were performed to determine the cause. Only subjects with a

fever of clinically undetermined origin were included in the

study, to avoid potential confusing factors. In the absence of

clinical evidence of complications, fever episodes were treated

with paracetamol alone. According to the WHO definition

(DeMayer et al. 1989), mild anaemia was defined as a haema-

tocrit of , 39%.

Statistical methods

Two population measures of parasitaemia were used: the

parasite rate, defined as the proportion of subjects with any

parasites in the smear at a given point in time; and the geo-

metric mean of parasite density (GMPD), defined as the

antilogarithm of arithmetic mean of the base 10 logarithm of

10 plus the parasite density.

Adjustment of parasite density by addition of the constant

10, needed to avoid infinities when taking logs, effectively

assigned the value 0.4 to an observed count of zero cells

infected out of the total of 300.

The associations of parasitaemia with time of day (0700,

1300, 1900 h) and day of study (day 1 to day 13) were exam-

ined through regression analyses conducted separately for the

dry and wet seasons. Logistic regression was used with the

parasite rate as the outcome measure and ordinary linear

regression with the parasitaemia index as the outcome.

Correlations between repeated measurements from each sub-

ject were accounted for by generalized estimating equation

(GEE) (Diggle et al. 1994). The independence model chosen

for the GEE ‘working correlation’ matrix meant that re-

gression coefficients could be estimated with standard pro-

grammes for (logistic) regression analysis of independent

data, while posthoc adjustments of standard errors were used

to account for the statistical dependence. The same approach

was taken to study the influence of parasite density on the

simultaneous occurrence of fever.

The association between the occurrence of at least one

© 2000 Blackwell Science Ltd406

fever episode and the maximum parasite density observed for

each individual during each season was tested for statistical

significance by the exact linear-by-linear association test

implemented in StatXact software. The records of subjects

who participated during both seasons were treated as statisti-

cally independent for this analysis.

Results

Thirty-nine people participated during the dry season and 41

during the wet season; 24 subjects participated during both

phases and 32 only once. No metabolites of chloroquine in

the urine were detected. Age does not appear as a significant

variable in the statistical analysis, because the subjects were

all adults. Slight variations observed in spleen size were also

not significant. Possible tuberculosis was diagnosed in one

participant on the first day of the dry season; he was ex-

cluded from the study. Sixty-three of the 79 subjects (80%)

presented with positive smears for malaria parasites at least

once. A mixed infection with P. falciparum and P. malariae

was found in six subjects. During the dry season, eight fever

episodes were recorded in six people; during the wet season,

eight episodes occurred in seven people. Table 1 summarizes

the distributions of the main study variables measured on the

first day of sampling.

Individual variations in parasitaemia

Profiles of parasite density as a function of time were exam-

ined for each individual. Although considerable intersubject

variation was observed, it was possible to identify four typical

patterns of the time evolution of parasitic density:

Type I: Negative smears during the entire follow-up

period.

Type II: Generally negative smears, but with one to three

peaks occurring without apparent periodicity.

Type III: Generally positive smears, with quite rare nega-

tives (Figure 1a,b,c).

Table 1 Summary data for the initial day of study

Dry season (n 5 38) Wet season (n 5 41)

GMPD* 194 573

Parasite rate† 042.1% 026.8%

Enlarged spleen‡ 010.5% 017.1%

Haematocrit, mean 6 041.8 6 4.4 041.0 6 3.6

Rate of anaemia§ 021.1% 026.8%

*Geometric mean parasite density among subjects with a positive

parasite count. †% of subjects infected. ‡Spleen measured the last

day. §Haematocrit , 39%.

Tropical Medicine and International Health volume 5 no 6 pp 404–412 june 2000

V. Delley et al. Single determination of malaria parasite density

© 2000 Blackwell Science Ltd 407

Type IV: Clearly marked periodicity, with episodes of

parasitaemia broken by periods of aparasitaemia. This

profile includes typical intermediate cases between types

2 and 3.

Others: Unclassifiable cases.

Figure 2 shows typical profiles of parasitaemia for four

study participants. The frequencies of the different profile

types are shown for both seasons in Table 2. Type III is the

most frequent profile for both seasons. Type I was next most

frequent in the dry season, and type IV in the wet season,

which is consistent with the higher transmission rate at this

time.

For the 24 subjects studied in both seasons, the profiles

differed from one season to the next, except for four indi-

viduals (two with type III on both occurrences, one with type

I, and one ‘other’). Two central observations from the para-

sitaemia profile analysis were that for many subjects the

parasite density varied substantially over the course of a few

hours; and that all subjects showed at least one negative

smear during the study.

For each season and each subject, ratios of parasite density

between pairs of samples taken 6 hours apart were calculated

thus: 24 ratios per individual (two per day) in the dry season

and 26 in the wet. Zero parasitaemia was recorded as 1 for

calculating ratios. Table 3 indicates for each season the num-

ber and proportion of individuals who had at least one and

three changes in parasite count by a factor of 100 and at least

one change by a factor of 1000 during the course of six

hours.

10

12

30

Day

(c)

GM

PD

2

Rainy season

25

20

15

4 6 8 10

35

12

60

Day

(d)

Par

asit

e ra

te

2

Rainy season

50

45

40

4 6 8 10

10

12

20

Day

(a)

GM

PD

2

Dry season

15

4 6 8 10

35

12

50

Day

(b)

Par

asit

e ra

te

2

Dry season

45

40

4 6 8 10

55

Figure 1 Fitted values (from regression models) of parasitaemia levels at 0700 h according to day of study. Solid circles represent unrestricted

day-to-day effects and solid lines the results of the corresponding quadratic regression model. (a) and (b) show the results for the dry season;

(c) and (d) for the wet season. (a) and (c) are for GMPD as the outcome measure while (b) and (d) are for parasite rate.

Table 2 Distribution of profiles by type

Dry season (n 5 38) Wet season (n 5 41)

————————— —————————

% n % n

Type I 28.9 (11) 12.2% 0(5)

Type II 07.9 0(3) 17.1% 0(7)

Type III 31.6 (12) 34.1% (14)

Type IV 15.8 0(6) 29.3% (12)

Others 15.8 0(6) 07.3% 0(3)

Tropical Medicine and International Health volume 5 no 6 pp 404–412 june 2000

V. Delley et al. Single determination of malaria parasite density

© 2000 Blackwell Science Ltd408

All subjects had at least one negative sample during the

13 days of the study. The 63 subjects who bore parasites dur-

ing this period had, on average, 14 negative samples (median

5 14) of 36 during the dry season, and 17 (median 5 15) of

39 during the wet season.

Variations in population measures of parasitaemia

The patterns of variation in parasitaemia at the individual

level were partly reflected in population averages calculated

for each of the 36 (dry season) or 39 (wet season) occasions

on which blood samples were drawn. The parasite rate (per-

centage of subjects infected at a given point in time) varied

between days, or between hours within a single day (28.9% to

57.9% during the dry season and 27.5% to 70.7% during the

wet season. The corresponding ranges for the GMPD were

26.8–64.1 and 29.3–138.5. These variations did not occur at

random. Parasite density was clearly elevated at midday com-

pared to morning or evening (Table 4). Examination of daily

averages demonstrated that, in both seasons, parasite density

first increased with day-of-study and subsequently declined.

The statistical significance of these effects was evaluated by

fitting regression models with time of day (hour) as a three-

level factor and day of study (day) either as a quadratic

35

41

Hours (total = 13 days)

(a)

Tem

per

atu

re (

°C)

7–1

5

log

10 scale

319271247223199175151127103795531 295

40

39

38

37

36

4

3

2

1

0

35

41

Hours (total = 13 days)

(b)

Tem

per

atu

re (

°C)

7–1

5

log

10 scale

319271247223199175151127103795531 295

40

39

38

37

36

4

3

2

1

0

35

41

Hours (total = 13 days)

(d)

Tem

per

atu

re (

°C)

7–1

5

log

10 scale

319271247223199175151127103795531 295

40

39

38

37

36

4

3

2

1

0

35

41

Hours (total = 13 days)

(c)

Tem

per

atu

re (

°C)

7–1

5

log

10 scale

319271247223199175151127103795531 295

40

39

38

37

36

4

3

2

1

0

Figure 2 Longitudinal profiles of four study participants as four types of example: (a) type I, (b) type II, (c) type III and (d) type IV. The solid

line represents the parasite density in relation to the sampling time (x-axis), the dotted line indicates the body temperature. The fever threshold

is shown as the horizontal line.

Dry season (n 5 38) Wet season (n 5 41)

————————– ————————

% (n) % (n)

At least one change by a factor of 100 or more 55.3 (21) 73.2 (30)

At least three changes by a factor of 100 or more 21.1 0(8) 41.5 (17)

At least one change by a factor of 1000 or more 13.2 0(5) 14.6 0(6)

Table 3 Variations of parasite density

between two samples, 6 h apart

Tropical Medicine and International Health volume 5 no 6 pp 404–412 june 2000

V. Delley et al. Single determination of malaria parasite density

© 2000 Blackwell Science Ltd 409

(second degree) polynomial, or as a factor with 12 or 13

levels, depending on the season. Adjusting for intrasubject

correlation increased the standard errors of the regression co-

efficients for day effects by a factor of two or three over those

obtained under the assumption of independence. Adjustment

decreased the standard errors for the hour effects by approxi-

mately the same amount. Wald tests, with two degrees of

freedom based on the adjusted variances, revealed that, for

the wet season, the hour and day effects were both statisti-

cally significantly correlated with either the parasite rate or

the GMPD (P # 0.001). For the dry season, in contrast, the

hour and the day effects, respectively, were statistically signifi-

cant (P 5 0.0005) only for the GMPD.

Fitted values for the daily parasitaemia levels at 0700 h

were calculated from the regression models and plotted

against day of study (Fig. 2). Despite the differences in levels

of statistical significance, the same general pattern was

observed in both seasons.

Association between fever and parasitaemia

There were 16 episodes of fever in 13 individuals. Thirteen

(81%) of the episodes occurred at the evening examination

(19 : 00 h) and none in the morning. The episodes were evenly

distributed between seasons (eight in each), making separate

evaluations of their association with parasitaemia inadvisable

owing to small numbers. The maximum risk was among sub-

jects with parasite densities . 1000 parasites/mm3, especially

. 10000 parasites/mm3, at least once during the follow-up

(P 5 0.009).Thus while there was a clear association between

fever and parasitaemia among individuals, it was more diffi-

cult to determine a relationship between the timing of the

Table 4 Average parasite density levels by season and time-of day

Measure Season 0700 h 1300 h 1900 h

Parasite rate (%) Dry 43.9 44.5 41.9

Wet 46.8 51.8 47.8

GMPD* Dry 42.6 48.1 38.4

Wet 48.1 63.7 46.7

*Geometric mean parasite density

35

41

Hours (total = 13 days)

(a)

Tem

per

atu

re (

°C)

7–1

5

log

10 scale

319271247223199175151127103795531 295

40

39

38

37

36

4

3

2

1

0

35

41

Hours (total = 13 days)

(c)

Tem

per

atu

re (

°C)

7–1

5

log

10 scale

319271247223199175151127103795531 295

40

39

38

37

36

4

3

2

1

0

Figure 3 (a) and (b) show two cases of high parasitaemia without fever. (c) and (d) show two subjects that would have been classified as apara-

sitaemic based on a single sample during a fever episode. Dotted line, temperature; solid line, parasitaemia.

35

41

Hours (total = 13 days)

(b)

Tem

per

atu

re (

°C)

7–1

5

log

10 scale

319271247223199175151127103795531 295

40

39

38

37

36

4

3

2

1

0

35

41

Hours (total = 13 days)

(d)

Tem

per

atu

re (

°C)

7–1

5lo

g10

sca

le

319271247223199175151127103795531 295

40

39

38

37

36

4

3

2

1

0

Tropical Medicine and International Health volume 5 no 6 pp 404–412 june 2000

V. Delley et al. Single determination of malaria parasite density

fever episode and the intraperson variations in parasite levels.

Individual profiles showed that only on a single occasion did

fever and the highest parasite density occur at the same time

during follow-up. Eleven of the 13 other fever cases observed

in Type II to IV profiles were separated from the time of

occurrence of the highest parasite density by an interval of

24–48 h. Two representative cases are shown in Figures 3c,d.

Figures 3a,b show examples of asymptomatic subjects with

parasite densities exceeding 15000 parasites/mm3. Logistic

regression analysis (GEE) of the probability that an indi-

vidual was febrile at a given point in time, evaluated as a

function of the concurrent parasite density, showed a positive,

but statistically non-significant, association.

Discussion

Individual variations in blood parasite density

This study in an endemic area confirms important fluctu-

ations in peripheral parasite density over the period of one

day in most asymptomatic subjects. An individual with

only 450 parasites/mm3 can reach a density of over 44000

parasites/mm3 only six hours later. The individual variations

observed in 48 h can go beyond 10000 parasites/mm3. This

exceeds by far the results predicted by the possible theoretical

multiplication of the parasites, which is approximately 10–20

times during a 48-h cycle (number of merozoites produced by

schizonts) (Kitchen 1949; Armitage & Blanton 1991). This

sudden increase in the observed peripheral parasitaemia can

be explained only by a massive release of sequestered para-

sitized erythrocytes.

Besides this sequestration phenomenon, synchronization

of the intraerythrocytic life cycle could partly explain the

amplitude of observed variations in parasitaemia. Both

mechanisms, sequestration-release and synchronization, are

responsible for important fluctuations in parasite density in

peripheral blood, and little is known about them in a natural

infection situation. In vitro, fever has a synchronizing effect

on the erythrocytic life cycle of the parasite (Kwiatkowski

1989; Kwiatkowski & Greenwood 1989). Variations in body

temperature can synchronize the different ‘broods of para-

sites’ (Hawking 1970). The fact that cycles were found only in

a part of the group or in some samples is not surprising, as

infection by P. falciparum is the least synchronous compared

with other plasmodia. The hypothesis that subjects who

developed fever are more likely to present ‘cyclical infections’

could not be tested because of the small number of registered

episodes.

The sequestration phenomenon is probably at the origin

of another remarkable observation: the frequent negative

smears, of which at least one was found in each of the in-

fected subjects. The transitory disappearance of the parasites

from the peripheral blood occurred at all times of day during

both seasons. These observations corroborate those of

Gendrel et al. (1992).

Variations in parasitaemia at the population level

The level of malaria in the group was estimated from the pro-

portion of subjects with detectable parasitaemia on examin-

ation of a blood smear, and from the GMPD. These two

measures are significantly different according to the time of

day (Table 4). Of the three daily samples, the one at 13 : 00

showed a significantly higher parasite density. These findings

agree with those of Hawking (1970), who showed that an

‘internal clock’ resulted in a synchronized explosion of

schizontes around noon.

Day of study also significantly modified the parasite den-

sity and the proportion of subjects harbouring parasites. This

was validated statistically using a quadratic regression model

(Fig. 2). The explanation for this effect is not readily appar-

ent. Technical biases were partially ruled out, as the smear

readings were done at the same time and by the same people

during the entire study. While one cannot rule out the influ-

ence of uncontrolled environmental factors (De Mello 1955),

this seems unlikely, as the effect was observed during both

seasons. Another possible hypothesis is that parasitaemia

shows cyclical variation. Moreover, chance effects cannot be

excluded.

Globally, these results show the limitations of classical

indices when they are used to gauge endemic malaria at a

particular point in time. The influence of the hour and the

day on the parasitaemia rate emphasizes the need to stan-

dardize sampling conditions.

Fever and parasitaemia

Although overall parasite counts were higher in subjects who

had a fever episode during follow-up, no temporal relation-

ship could be established between these two conditions.

Fifteen of 16 episodes of fever did not coincide with a para-

sitaemia peak. While the limited number of fever cases man-

dates caution in drawing conclusions, the longitudinal data

did allow us to investigate the temporal association between

parasite density and fever. The observed profiles are com-

patible with our third and fourth hypotheses, which state that

the parasite density needed to trigger fever strongly varies

from one subject to another (Rogier et al. 1996) even in

adults. In fact, we encountered subjects with fever episodes

whose profiles ranged from one with nearly no parasites to

another with consistently high levels.

As immunity develops, the number of parasites an organ-

ism can tolerate before developing fever increases (Rogier &

Trape 1993; Smith et al. 1994b). Density thresholds have been

© 2000 Blackwell Science Ltd410

Tropical Medicine and International Health volume 5 no 6 pp 404–412 june 2000

V. Delley et al. Single determination of malaria parasite density

proposed for the development of fever in response to malaria

infection. In 1958, Miller proposed a threshold of 500

parasites/mm3 for adults. Today, the thresholds vary accord-

ing to author and study population (endemic zone, season,

adults, children) from 500 to 15000 parasites/mm3 (Trape

et al. 1985; Baudon et al. 1986; Velema et al. 1991)

In our study, close to one-third of infected subjects reached

the threshold of parasite density defined by Miller (1958) on

at least one occasion without developing any symptoms. This

contradicts the widespread notion that it is rare to find very

high asymptomatic parasitaemia in adults. We observed para-

sitaemia that fell from over 25000 to a very low level in the

absence of clinical symptoms. The extreme variability of the

measured peripheral parasitaemia could be responsible, in

part, for the different threshold levels proposed in the litera-

ture, and leads us to question the very concept of a parasite

density threshold in ascertaining the malaria origin of a fever

access.

Validity of a single blood sample for assessing the causal

relationship between parasitaemia and fever

It is probable that the total quantity of parasites present in

the organism largely determines the importance of clinical

symptoms and pathological manifestations (Armstrong-

Schellenberg et al. 1994). However, the peripheral measure-

ment of the parasite density does not represent the total

parasitic load. A single peripheral blood sample cannot

validly establish or refute the malarial origin of a fever

episode. For these reasons the interpretation of parasite den-

sities, especially when measured only once, demands great

care both at an individual level, in estimating the total para-

site load, and at an epidemiological level for estimating the

population parasitic infection rate.

In the natural situation, peripheral variations in parasitic

density appear to be very complex. This study demonstrated

the influence of time of day and day of sampling. More in-

depth research is necessary, however, to understand the other

factors that may modify peripheral parasitaemia. A longer

period, notwithstanding the resulting ethical and organiz-

ational problems, is needed to model variations in para-

sitaemia. In a previous longitudinal study on children from

the same village (Bouvier et al. 1997), we observed that the

value of parasite density for predicting a fever occurrence was

lower than that of other indicators such as season, age or

haematocrit during the dry season. Another 10-day study of

children during the wet season with initial parasitaemia above

10000 parasites/mm3, showed a fever incidence barely higher

than that of children with much lower or no initial para-

sitaemia (Delley 1998).

Our results show again that parasitaemia remains but one

among a multitude of elements necessary to establish a pre-

cise causal diagnosis of malarial fever in an endemic area. We

believe that the proof of the malarial origin of a fever episode

must combine detection of blood parasites (when possible) in

the framework of a structured decision process, as outlined

by Rougemont et al. (1991).

Acknowledgements

Thanks to Mrs. Pauline Duponchel in Bamako for her valu-

able assistance, the team of interviewers in Sikasso, the health

team of Bougoula, the District Medical Officers of the Cercle

de Sikasso and the Regional Direction of Public Health for

their kind collaboration. We thank the population of

Bougoula for its support of this project. Supported by a grant

from the National Fund for Scientific Research, Switzerland

(No 3200-037810.93).

References

Armitage KB & Blanton RE (1991) Paroxysmal fluctuations in

observed parasitemia in Plasmodium falciparum malaria (letter).

American Journal of Medicine 90, 530–531.

Armstrong-Schellenberg JRM, Smith T, Alonso PL & Hayes RJ

(1994) What is clinical malaria? Finding case definitions for field

research in highly endemic areas. Parasitology Today 10, 439–442.

Baudon D, Gazin P, Sanou JM et al. (1986) Morbidité palustre en

milieu rural au Burkina Faso; étude de 526 cas fébriles. Médecine

D’afrique Noire 33, 767–776.

Bouvier P, Doumbo O, Breslow N et al. (1997) Seasonality, malaria

and impact of prophylaxis in a West African village: I. Effect on

anaemia in pregnancy. American Journal of Tropical Medicine and

Hygiene 56, 378–383.

Bouvier P, Rougemont A, Breslow N et al. (1997) Seasonality and

malaria in a West African village: does high parasite density pre-

dict fever incidence? American Journal of Epidemiology 145,

850–857.

Bruce-Chwatt LJ (1963) A longitudinal survey of natural malaria

infection in a group of West African adults. II. West African

Medical Journal 12, 199–217.

De Mello JP (1955) The influence of atmospheric changes on the

relationship of the malaria parasite and the human host. East

African Medical Journal 32, 137–140.

Delley V (1998) Approche épidémiologique du paludisme en zone

d’endémie: incidence de la fièvre chez des enfants de 1–12 ans et

suivi des variations naturelles de la parasitémie chez des adultes.

PhD Thesis, Université de Genève.

DeMayer EM, Dallman P, Gurney JM et al. (1989). Preventing and

Controlling Iron Deficiency Anaemia Through Primary Health

Care. WHO, Geneva.

Diggle PJ, Liang KY & Zeger S (1994). Analysis of Longitudinal

Data. Calendon Press, Oxford.

Farnert A, Snounou G, Rooth I & Bjorkman A (1997) Daily dynam-

ics of Plasmodium falciparum subpopulation in asymptomatic

children in a holoendemic area. American Journal of Tropical

Medicine and Hygiene 56, 538–547.

© 2000 Blackwell Science Ltd 411

Tropical Medicine and International Health volume 5 no 6 pp 404–412 june 2000

V. Delley et al. Single determination of malaria parasite density

Adaptation of the Sakers–Solomons test: simple, reliable colori-

metric field assays for chloroquine and metabolites in urine.

Bulletin of the World Health Organization 67, 295–300.

Rogier C, Commenges D & Trape JF (1996) Evidence for an age-

dependent pyrogenic threshold of Plasmodium falciparum para-

sitaemia in highly endemic populations. American Journal of

Tropical Medicine and Hygiene 54, 613–619.

Rogier C & Trape JF (1993) Malaria attacks in children exposed to

high transmission: who is protected? Transactions of the Royal

Society of Tropical Medicine and Hygiene 87, 245–246.

Rougemont A, Breslow N & Brenner E (1991) Epidemiological basis

for clinical diagnosis of childhood malaria in endemic zones in

West Africa. Lancet 338, 1292–1295.

Smith T, Genton B, Baea K et al. (1994b) Relationships between

Plasmodium falciparum infection and morbidity in a highly

endemic area. Parasitology 109, 539–549.

Smith T, Schellenberg J & Hayes R (1994a) Attributable fraction esti-

mates and case definitions for malaria in endemic areas. Statistics

in Medicine 13, 2345–2358.

Toure YT, Dolo G, Petrarca V et al. (1998) Mark–release–recapture

experiments with Anopheles gambiae s.l. in Banambani Village,

Mali, to determine population size and structure. Medical

Veterinary Entomology 12, 74–83.

Trape JF, Peelman P & Morault-Peelman B (1985) Criteria for diag-

nosing clinical malaria among a semi-immune population exposed

to intense and pernial transmission. Transactions of the Royal

Society of Tropical Medicine and Hygiene 79, 435–442.

Velema JP, Alihonou E, Chippaux JP et al. (1991) Malaria morbidity

and mortality in children under three years of age on the coast of

Benin, West Africa. Transactions of the Royal Society of Tropical

Medicine and Hygiene 85, 430–435.

© 2000 Blackwell Science Ltd412

Gendrel D, Kombila M, Martz K et al. (1992) Parasitémie au cours

des accès palustres à Plasmodium falciparum chez l’enfant. Presse

Médicale 21, 1805–1808.

Genton B, Smith T, Baea K et al. (1994) Malaria: how useful are clini-

cal criteria for improving the diagnosis in a highly endemic area?

Transactions of the Royal Society of Tropical Medicine and

Hygiene 88, 537–541.

Gilles HM (1993) Epidemiology of malaria. In: Bruce-Chwatt’s

Essential Malariology. 3rd edn (eds HM Gilles & DA Warrell),

Little, Brown and Co., Boston, pp. 124–162.

Greenwood BM, Bradley AK, Greenwood AM et al. (1987) Mortality

and morbidity from malaria among children in a rural area of the

Gambia, West Africa. Transactions of the Royal Society of

Tropical Medicine and Hygiene 81, 478–486.

Hawking F (1970) The clock of the malaria parasite. Science America

Journal (June), 222, 123–131.

Hogh B, Marbiah NT, Peterson E et al. (1993) Classification of clini-

cal falciparum malaria and its use for the evaluation of chemo-

suppression in children under six years of age in Liberia, West

Africa. Acta Tropica 54, 105–115.

Kitchen SF (1949) Malariology, Vol. 2 (ed. MF Boyd), WB Saunders,

Philadelphia, pp. 966–1045.

Kwiatkowski D (1989) Febrile temperatures can synchronize the

growth of Plasmodium falciparum in vitro. Journal of Experi-

mental Medicine 169, 357–361.

Kwiatkowski D & Greenwood BM (1989) Why is malaria fever per-

iodic? A hypothesis. Parasitology Today 5, 164–166.

Miller MJ (1958) Observation on the natural history of malaria in

the semi-resistant West-Africa. Transactions of the Royal Society

of Tropical Medicine and Hygiene 52, 152–168.

Mount DL, Nahlen BL, Patchen LC & Churchill FC (1989)