Hemoglobin C is associated with reduced Plasmodium falciparum parasitemia and low risk of mild malaria attack.
ABSTRACT Genetic predisposition to malaria has been shown by epidemiological, case-control and linkage studies. In particular, case-control studies have recently shown association between hemoglobin C and resistance to severe malaria in Mali and to clinical malaria in Burkina Faso. In a longitudinal study of families living in an endemic area, we investigated whether hemoglobin C is associated with reduced Plasmodium falciparum parasitemia and low risk of mild malaria attack. We surveyed 256 individuals (71 parents and 185 sibs) from 53 families during 2 years. Hemoglobin C carriers had less frequent malaria attacks than AA individuals within the same age group (P=0.01). Since age correlated with malaria attack and parasitemia (P<0.0001), we took age into account in association analyses. We performed combined linkage and association analyses, which avoid biases due to population structure. Using multi-allelic tests, we evidenced association between hemoglobin genotype and phenotypes related to malarial infection and disease (P<0.001). We further analyzed individual hemoglobin alleles and detected negative association between hemoglobin C and malaria attack (P=0.00013). Analyses that took into account confounding factors confirmed the negative association of hemoglobin C with malaria attack (P=0.0074) and evidenced a negative correlation between hemoglobin C and parasitemia (P=0.0009). These associations indicate that hemoglobin C reduces parasitemia and confers protection against mild malaria attack.
- SourceAvailable from: nih.govBritish medical journal 11/1963; 2(5363):976-8.
- [show abstract] [hide abstract]
ABSTRACT: Cross-sectional and longitudinal studies were performed in a rural population living in The Gambia to examine the relationship between several in vitro assays of the host immune response to asexual stages of Plasmodium falciparum and protection from malaria in vivo. Assays included an enzyme-linked immunosorbent assay for antibodies to schizont antigens; an indirect immunofluorescence assay for total antiblood-stage antibodies; an immunofluorescence assay on glutaraldehyde-fixed parasites to detect antibodies to antigen Pf 155; an assay for serum inhibition of red blood cell invasion; a micro-agglutination assay to detect antibodies to neo-antigens on the surface of infected red blood cells; and an assay using polymorphonuclear leucocytes to detect antibodies capable of opsonizing schizont infected red blood cells. There were marked differences in the age-related pattern of response for different assays performed on sera obtained at a cross-sectional survey of 280 individuals. Examination of the correlation between the various immune responses and malariometric indices at the population level and at the individual level provided no evidence that any of the in vitro assays were related to protective immunity. The relationship between in vitro measurements of the anti-malarial immune response and protection from clinical episodes of malaria was examined in a group of 134 children aged 11 years and under who were monitored weekly throughout an entire malaria transmission season. The only immune factor to show a consistent protective effect against clinical malaria was the titre of antibodies to neo-antigens on the infected erythrocyte surface (P = 0.01). The same longitudinal techniques were used to examine the effect of two non-immunological factors, sickle cell trait and mosquito net usage, both of which showed significant protection against clinical episodes and malaria.Transactions of the Royal Society of Tropical Medicine and Hygiene 01/1989; 83(3):293-303. · 1.82 Impact Factor
- [show abstract] [hide abstract]
ABSTRACT: Plasmodium falciparum malaria remains a major cause of morbidity and mortality in many tropical countries, especially those in sub-Saharan Africa. Human genetic control of malaria infection is poorly understood; in particular, genes controlling P. falciparum blood infection levels remain to be identified. We recently evidenced the existence of complex genetic factors controlling blood infection levels in an urban population living in Burkina Faso. We performed, on 153 sibs from 34 families, sib-pair linkage analyses between blood infection levels and chromosome 5q31-q33, which contains numerous candidate genes encoding immunological molecules. Our results, obtained by means of the two-point Haseman-Elston (HE) method and a nonparametric (NP) approach, show linkage of parasitemia to D5S393 (P=.002) and D5S658 (P=.0004). Multipoint analyses confirmed linkage, with a peak close to D5S658 (P=.0013 and P=.0007 with the HE and NP methods, respectively). The heritability of the locus was .48, according to the two-point results, and .43, according to the multipoint results; this indicates that its variation accounted for approximately 45% of the variance of blood infection levels and that the locus plays a central role in the control of parasitemia. The identification of the gene is, therefore, of major interest in understanding the mechanisms controlling P. falciparum parasitemia.The American Journal of Human Genetics 09/1998; 63(2):498-505. · 11.20 Impact Factor
Hemoglobin C is associated with reduced
Plasmodium falciparum parasitemia and
low risk of mild malaria attack
Pascal Rihet1,*, Laurence Flori1, Franc ¸ois Tall2, Alfred S. Traore ´2and Francis Fumoux1
1Universite ´ de la Me ´diterrane ´e, Marseille, France and2Universite ´ de Ouagadougou, Ouagadougou, Burkina Faso
Received May 28, 2003; Revised October 9, 2003; Accepted October 24, 2003
Genetic predisposition to malaria has been shown by epidemiological, case–control and linkage studies. In
particular, case–control studies have recently shown association between hemoglobin C and resistance to
severe malaria in Mali and to clinical malaria in Burkina Faso. In a longitudinal study of families living in an
endemic area, we investigated whether hemoglobin C is associated with reduced Plasmodium falciparum
parasitemia and low risk of mild malaria attack. We surveyed 256 individuals (71 parents and 185 sibs) from
53 families during 2 years. Hemoglobin C carriers had less frequent malaria attacks than AA individuals
within the same age group (P¼0.01). Since age correlated with malaria attack and parasitemia (P<0.0001), we
took age into account in association analyses. We performed combined linkage and association analyses,
which avoid biases due to population structure. Using multi-allelic tests, we evidenced association between
hemoglobin genotype and phenotypes related to malarial infection and disease (P<0.001). We further
analyzed individual hemoglobin alleles and detected negative association between hemoglobin C and
malaria attack (P¼0.00013). Analyses that took into account confounding factors confirmed the negative
association of hemoglobin C with malaria attack (P¼0.0074) and evidenced a negative correlation between
hemoglobin C and parasitemia (P¼0.0009). These associations indicate that hemoglobin C reduces
parasitemia and confers protection against mild malaria attack.
The complex genetic control in human malaria reflects
thousands of years of selective pressure. Hemoglobinopathies
were the first polymorphisms thought to be selected under the
pressure of malaria. On the basis of the worldwide distribution
of hemoglobin S (HbS) and a-thalassemia, Haldane (1) and
Allison (2) suggested that a-thalassemia and HbS gave a
selective advantage for survival in malaria-endemic areas.
Case–control studies have shown the association of resistance
to severe malaria with both a-thalassemia (3) and HbS (4,5).
Case–control studies have been performed to investigate the
role of other candidate genes in severe malaria. Associations
have been found between resistance and genes encoding other
red blood cell proteins (Band 3, G6PD) or immunological
molecules (HLA-B, HLA-DR, TNFa, ICAM1, CD36, iNOS
and IFNR) (6,7). It should be stressed that some associations
have not been confirmed, suggesting that genetic control in
human malaria may differ between populations. Another
explanation would be that population associations may occur
in the absence of linkage as a result of admixture, heterogeneity
or stratification in a population. In this case, association
between genes and resistance to malaria does not mean that the
candidate genes associated are involved in human malaria. To
circumvent the problem of population structure, linkage and
association analyses should be performed (8). Such methods
have been successfully used to analyze genetic control of
parasitemia (9,10) and of mild malaria (11,12).
Strikingly, most of the genes associated with resistance to
severe malaria are not associated with mild malaria or para-
sitemia (6,7). Malaria pathogenesis is incompletely known and
the identification of genes controlling phenotypic variations
related to malaria infection should be helpful in understanding
the mechanisms involved. In particular, this may clarify the
relationship between parasitemia and clinical malaria, and
between mild malaria and severe malaria.
Several case–control studies have tested the association
between hemoglobin C (HbC) and resistance to clinical
*To whom correspondence should be addressed at: Universite ´ de la Me ´diterrane ´e, IFR48, Faculte ´ de Pharmacie, Laboratoire d’Immunoge ´ne ´tique et de
Pharmacologie du Paludisme-EA 864, 27 Bd Jean Moulin, 13385 Marseille Cedex 5, France. Tel/Fax: þ33 491803674; Email: email@example.com
Human Molecular Genetics, 2004, Vol. 13, No. 1
Advance Access published on November 12, 2003
Human Molecular Genetics, Vol. 13, No. 1 # Oxford University Press 2004; all rights reserved
by guest on June 1, 2013
malaria. Results from a study in Nigeria and one in Mali
indicated lack of protection (13,14), while association between
heterozygosity for C and protection against severe malaria was
detected in another study in Mali (15). Recently, a large case–
control study in Burkina Faso detected a strong association
between resistance to clinical malaria and HbC in both the
heterozygous and the homozygous state (16). However, case–
control or cross-sectional studies did not detect association of
HbC with parasitemia (15–17).
The aim of the present study was to test association in
the presence of linkage between hemoglobin genotype on the
one hand and parasitemia and risk of mild malaria attack on the
other hand. A longitudinal study was conducted over 2 years on
256 subjects from 53 families living in an urban area in
Burkina Faso. We report here the results of family-based associ-
Hemoglobin genotypes and phenotypes were available for 256
individuals (71 parents and 185 sibs) belonging to 53 families
with between two and nine members. The frequencies of
hemoglobin A, C and S were respectively 0.814, 0.139 and
0.047 for all individuals. AA, AS, AC and CC genotype freq-
uencies were in Hardy–Weinberg equilibrium (P¼0.45). For
sibs, Table 1 shows the age distribution among the different
hemoglobin groups. The age distributions of AA sibs and sibs
with HbC were similar (w2¼2.1, df¼4, P¼0.71). Table 2
shows the frequencies of genotype AA, AS, AC and CC in
affected sibs. Forty-seven percent of AA sibs, 22% of AS sibs
and 25% of AC sibs presented at least one malaria attack (P1)
during the study: hemoglobin genotypes, therefore, appeared to
influence the occurrence of malaria attack (w2¼9.1, df¼2,
P¼0.01). Figure 1 shows that HbC carriers had less frequent
malaria attacks than AA individuals belonging to the same age
class. Since the number of children in the youngest age class
was low (Table 1), children aged 1–5 years were excluded in
some analyses. The logistic regression, which included the
6–10, 11–15, 16–20 and 21–25 age classes as explanatory
variable, showed that the association between HbC and the
occurrence of malaria attack was significant (w2¼6.1, df¼1,
P¼0.013). When children aged 1–5 years were included in the
analysis, the negative association of HbC with malaria attack
was also found (w2¼6.2, df¼1, P¼0.012). The logistic regres-
sion treating age as a continuous variable led to very similar
results (w2¼5.8, df¼1, P¼0.016). The odds of malaria attack
between AA individuals and HbC carriers was 2.7 (95%
confidence interval 1.2–6.1).
Age correlated with the four phenotypes: malaria attack (P1),
risk of developing malaria attack (P2), mean of adjusted
parasitemia (P3) and maximum parasitemia (P4) (Table 3).
Fifty-four percent (7/13) of sibs 0–5 years of age, 60% (28/47)
of those 6–10 years of age, 40% (22/55) of those 11–15 years of
age, 35% (14/40) of those 16–20 years of age and 7% (2/30) of
those 21–25 years of age had at least one malaria attack (P1)
during the study. As expected, age influenced the occurrence
of malaria attack (w2¼22.9, df¼4, P¼0.00013). Logistic
regression considering age as explicative variable on the prob-
ability of malaria attack (P2) also showed a highly significant
effect of age in the sample of 185 sibs (Table 4). Age also had
an effect on mean adjusted parasitemia (P3) and maximum
parasitemia (P4) in the sample of 185 sibs (Table 4).
Polynomial regression analysis showed that the best fitting
function was linear. The regression lines accounted for 14 and
21% of the variance of P3 and P4, respectively. Hence, age was
retained for combined linkage and association analyses.
Table 5 shows the results of multi-allelic tests for binary and
quantitative phenotypes. Hemoglobin was associated with P1,
P2, P3 and P4. The FBAT results showed a deficit transmission
of HbC from parents to the offspring with malaria attack. HbC
was negatively associated with malaria attack (Z¼?3.82;
P¼0.00013). In the same analysis, HbA was positively asso-
ciated with malaria attack (Z¼4.27; P¼0.00002). Similar
analyses with HbS did not evidence a significant association
with malaria attack. The trend, however, was negative, and we
lumped together HbS and HbC in some association analyses.
We evidenced negative association of grouped alleles S and C
with malaria attack (Z¼?3.94; P¼0.000082).
We also tested the association between individual hemo-
globin alleles and quantitative phenotypes. Table 6 shows
the QTDT results for phenotypes, for which we evidenced
association using multi-allelic tests. HbC was found to be
negatively associated with P2 (P¼0.0074), P3 (P¼0.0012)
and P4 (P¼0.0009) (Table 6). We also found a trend in favor
of a protective effect of HbS, although it was not significant.
Since our data suggested that both HbS and HbC protect
against malaria, we grouped alleles S and C in further analyses.
We further found evidence of associations with P2 (P¼
0.0007), P3 (P¼0.0001) and P4 (P¼0.0003).
In this study, we used family-based association tests to
investigate the association between HbC on the one hand and
parasitemia and malaria attack on the other hand. Several
phenotypes related to malaria infection and disease were tested
for association with HbC.
Thirty-five percent of the sibs were carriers of either HbS or
HbC. The gene frequency of HbC was higher than the
frequency of HbS. This is consistent with the high frequency of
HbC in West Africa (15,17,18). The frequencies of AA, AS and
AC individuals reported in central Burkina Faso by Modiano
et al. (16) are very close to the frequencies we report here in a
population in the West of Burkina Faso.
Raw data indicated that AS and AC individuals less frequently
develop mild malaria attack than AA individuals. The protective
effect of HbS on severe malaria (4) and on mild malaria (19,20)
was previously reported. Association of HbC with severe
malaria has been found in Mali (15) and association of HbC
with clinical malaria (severe and mild malaria) has been found
in Burkina Faso (16). In the latter study, no difference in HbC
was observed between severe and mild malaria, suggesting that
HbC also protects against mild malaria. To address this issue,
we performed family-based association analyses. We evidenced
a strong association between HbC and protection against mild
malaria (P1). In our study area, the occurrence of malaria attack
slightly decreased with increasing age in individuals less than
20 years of age, as previously described in areas with similar
2Human Molecular Genetics, 2004, Vol. 13, No. 1
by guest on June 1, 2013
malaria transmission intensities (21). The influence of age was,
nevertheless, highly significant. We, therefore, took age into
account in further analyses and we confirmed the association
between HbC and protection against mild malaria (P2).
Since high parasitemia strongly enhances the risk of malaria
attack, we further analyzed the association between Hb
genotypes and mean of adjusted parasitemia (P3). We found
a negative association between HbC and P3. Our findings
contrasted with case–control and cross-sectional studies, which
showed no correlation of HbC heterozygosity with parasite
rates and densities (15–17). It should be stressed, that we report
here a longitudinal and familial study. Because of the
fluctuation of parasitemia, longitudinal studies are likely
required to address this issue.
Studies that evaluated association between HbS heterozy-
gosity and parasite density also showed contrasting results:
parasite density in AS individuals was lower (2,17,19), similar
(22) or higher (23) when compared with parasite density in AA
individuals. It was proposed that HbS may limit the expansion
of malarial infection at high parasitemia, since AS individuals
with high parasitemia were rarely found (5,17,18). A similar
effect of HbC was also suggested (17,18). In this case, the
search for maximum parasitemia in a longitudinal survey may
We therefore designed another phenotype that may better
reflect potential bursts of parasite multiplication: maximum
parasitemia (P4). P4 was a quantitative phenotype and
represented the highest parasitemia in each individual. Raw
data suggested that AA individuals may develop higher
maximum parasitemia than did AS, AC and CC individuals.
We took into account the influence of confounding factors,
such as age, in further analyses. Results of multi-allelic QTDT
showed a clear association between hemoglobin genotypes and
maximum parasitemia (P4). HbC was negatively associated
with maximum parasitemia.
These results clearly indicate that parasite expansion is
inhibited in individuals with HbC. This is consistent with
in vitro studies showing lower parasite multiplication rates in
CC than in AA red cells (24–26). These abnormal cells
probably represent a barrier for the parasite because of their
inability to lyse and release merozoites at the appropriate stage
(25). In addition, ring forms and trophozoites showed evidence
of disintegration within a subset of CC red cells (26). Although
AC red cells sustain normally the growth of P. falciparum in
vitro (24,25), our findings clearly show a protective effect of
heterozygosity for HbC against the parasite. It seems very
likely that the in vivo protective effect of HbC depends on
factors, which are not in the in vitro culture system. In
particular, the protective effect of HbC may act in synergy with
specific acquired immunity, as suggested for the protective
effect of HbS (19,27,28).
Whatever the mechanisms involved, we propose that the
inhibitory effect of HbC on parasitemia may partly explain
the protective effect of HbC on mild malaria. Carriers of HbC,
who have a reduced parasitemia, would present a diminished
risk of mild malaria attack. One might also assume that the
protective effect of HbC against the parasite may participate in
the protective effect of HbC against severe malaria.
In our study, HbS was not significantly associated with
reduced parasitemiaand low
Nevertheless, the frequency of HbS was too low to detect a
significant association and our results are not in contradiction
to previous reports (19,20). Furthermore, we found a negative
trend in favor of a protective effect of HbS against malaria, and
we evidenced association of grouped alleles S and C with
reduced parasitemia and low risk of malaria attack.
In conclusion, the main results of the family-based associa-
tion analyses showed that HbC is associated with low
parasitemia and protection against mild malaria. Our results
are in line with those of others who have well established the
protection afforded by HbS. Strikingly, the HbC and the HbS
mutations occur in the 6th position in the b-globin DNA
sequence, and both mutations coexist in several populations in
West Africa. Since HbC appears to have few adverse effects,
compared with HbS in the homozygous state, it has been
suggested that HbC would replace HbS in West Africa (17).
MATERIALS AND METHODS
The study subjects live in an urban district of Bobo-Dioulasso,
the second largest town of Burkina Faso, in an endemic area for
malaria. P. falciparum transmission occurs only during the
rainy season (August–December). During the 2 years of
the study, the entomological inoculation rate was close to
Table 1. Age distribution of Hb genotype
Age1 to 5 6 to 1011 to 15 16 to 2021 to 25 All
AC and CC
Age distribution is given for sibs. CC individuals were 9 and 24 years old.
Table 2. Clinical data of sibs
aSibs with at least one malaria attack during the 24 months of the study.
Human Molecular Genetics, 2004, Vol. 13, No. 13
by guest on June 1, 2013
30 infective bites per year. The study population and the area of
parasite exposure have been described elsewhere (29). The
study included 256 individuals (71 parents and 185 sibs) from
53 families; the mean age of the sibs was 13.0?5.4 (range 1–
24). Most of the population belongs to the Mossi ethnic group
(50%); the other groups included the Dafing (19%), Guian
(5%), Bissa (15%), Samogo(1%), Bobo (6%) and Nounouma
(4%). The seasonal intensity of parasite transmission was
homogeneous within the area. The whole population volun-
teered to participate in the study, and all participants were
clearly informed of the objective and the protocol. The Expert
Committee of the University of Ouagadougou and the Medical
Authority of Burkina Faso approved the study protocol.
Determination of parasitemia was described in our previous
study (30). Briefly, each family was visited 20 times during the
24 months of the study and parasitemia was measured. In
addition, parasitemia was measured during febrile episodes.
The mean number of parasitemia measurements per subject
was 13.4?5.5 (range 1–25). Fingerprint peripheral blood
samples were taken from all family members present and thick
and thin blood films were stained with Giemsa. The parasite
determination and numeration were established blindly from
two independent readings. Only P. falciparum asexual forms
were retained to determine parasitemia. Parasitemia was
defined as the number of parasitized erythrocytes observed
per ml in thin blood films.
Febrile episodes were extensively recorded by active case
detection over 24 months. For patients with fever, a thick blood
film was prepared by the standard procedures. Diagnosis of
mild malaria attack was based on P. falciparum parasitemia,
fever (axillary temperature more than 37.5?C) and clinical
symptoms (headache, aching, vomiting or diarrhea in children);
in that case no threshold of parasitemia was used. In the
absence of classical symptoms of malaria, and once other
pathologies could not be eliminated, only children (age <15
years) with more than 5000 parasites per ml and older subjects
with more than 2000 parasites per ml were considered as having
had a malaria attack. According to the recommendation of the
CNRFP (Centre National de Recherche et Formation sur le
Paludisme) of Burkina Faso, each episode of illness was treated
with 25mg/kg chloroquine over 3 days or until recovery.
Parasitemia was checked at the end of the treatment. Eighty
individuals (seven parents and 73 sibs) presented at least one
mild malaria attack during the survey. No cases of severe
malaria were recorded.
Determination of phenotypes
Table 3 describes the phenotypes used in the study.
Mild malaria attack (P1) was directly based on clinical data.
Subjects who presented at least one mild malaria attack during
the survey were considered affected. The others were con-
The second phenotype (P2) was based on the risk of
developing malaria attacks. To take into account the influence
of covariates on mild malaria attack, we performed logistic
regression. Age was considered a continuous variable. The
logit of the probability P of malaria attack can be expressed in
the form of function of age as follows: log(P/17P)¼
b0 þ baAge, where b0 is constant and ba is the regression
coefficient (31). The residual of the logistic regression model,
which considered age a continuous variable (P2) was used in
association and linkage analyses. In some analyses, age was
also categorized into five classes.
Mean of adjusted parasitemia (P3) was a logarithmic
transformation of the parasitemia adjusted for seasonal trans-
mission (30). All the parasitemia data were included. To take
into account the seasonality of the transmission, the influence
of the date of the visits on ln(1 þ parasitemia) (LP) was
evaluated by one way analysis of variance. The mean LP
observed during each visit was calculated. The individual LP
was then corrected for the visit effect by subtracting from each
individual LP the mean LP of the corresponding visit. Multiple
polynomial regression was done with age and the number of
measurements. The explicative variables were treated as conti-
nuous variables. The number of measurements did not
correlate with P3. In contrast, age strongly correlated with
P3 and was retained as a covariate for linkage and association
Maximum parasitemia (P4) was based on a logarithmic
transformation of the highest parasitemia in each individual.
Multiple polynomial regression was done as indicated for P3.
The analysis revealed that age and the number of measurements
had an effect on P4 (P¼0.0001). We took into account age
and the number of measurements in association and linkage
All computations were done with SPSS software (SPSS,
Boulogne, France). Tests of the hypothesis that a regression
coefficient was 0 were done with the Wald w2statistic for logistic
regression and Student’s t-test for multiple polynomial regression.
Only terms significant at the 5% level were retained.
Figure 1. Percentage of malaria attack according to age in AA group (u)
(n¼114) and in AC and CC group (j) (n¼41). AA (n¼7) and AC sibs
(n¼5) aged less than 5 years were not included. The inclusion of the sibs aged
1–5 years in the analysis did not alter the distribution. Within the 1–10 age
class, 67.6% of AA individuals and 43.7% of HbC carriers were affected.
4Human Molecular Genetics, 2004, Vol. 13, No. 1
by guest on June 1, 2013
Blood samples were taken from 256 individuals by venipunc-
ture. The hemoglobin genotypes were identified by electro-
phoresis of red blood cell lysates on acetate membrane at an
alkaline pH. Acetate sheets were stained with ponceau red,
yielding discrimination of hemoglobins A, S and C.
Combined linkage and association analysis
Association in the presence of linkage was assessed using
family-based tests, which avoid biases due to population
stratification, population heterogeneity, or population admix-
ture, and are designed to deal with multiple alleles.
Association and linkage of binary trait were assessed using
the FBAT program (32). The FBAT statistics, which uses data
from sibships in nuclear families takes into account sibling
correlations. The default null hypothesis tested is no linkage
and no association. The statistics under this hypothesis
calculated under the distribution of offspring genotype are
conditional on parental genotype and on trait values for
offspring and parents. FBAT calculates a Z score and a 2-side
P value based on a normal approximation.
Combined association and linkage analyses of quantitative
traits were carried out using the orthogonal model released in
the QTDT 2.3.0 program (33). Variance components are used
to construct a test that utilizes information from all available
offspring. It is a general linkage-disequilibrium test that is
applicable to the analysis of quantitative traits in nuclear
families of any size, with or without parental genotypes.
Evidence of association can be evaluated by likelihood-ratio
test (null hypothesis likelihood L0 versus alternative hypoth-
esis likelihood L1). Asymptotically, the quantity 2(lnL1-lnL0)
is distributed as w2with df equal to the difference in number
of parameters estimated. Because variance components can
be sensitive to the phenotypic distribution, we checked the
hypothesis of multivariate normality with the SPSS software
(SPSS, Boulogne, France). Violation of multivariate normality
warrants the Monte-Carlo permutation test. In this case, we
performed 9999 permutations to calculate empirical P values.
We acknowledge all the volunteer families of Bobo Dioulasso.
We thank Y. Traore ´, T. Traore ´-Leroux and the entomological
team for their contributions, and G. Milon for helpful dis-
cussion during the study. The work was supported by the
French Ministry of Research and Technology, the Fondation
Table 6. Association of alleles with quantitative phenotypes
Direction of association
P4 (Age, number of measurements)
aCovariates that influence the phenotypes are taken into account using QTDT.
bEmpirical P values for non-normal data.
Table 3. Phenotypes related to malarial infection and disease
Phenotype Type of phenotype CovariatesDistribution of the residual
Malaria attack (P1)
Risk of developing malaria attacks (P2)
Mean of adjusted parasitemia (P3)
Maximum parasitemia (P4)
Age, number of measurements
Table 4. Correlationaof P2, P3, P4 with ageb
aThe analysis included only sibs.
bAge was treated as a continuous variable in each model.
cLogistic regression was used; the P value was calculated by using the Wald w2
dLinear regression was used; the P value was calculated by using Student’s t-test.
Table 5. Multi-allelicatestsbof association for hemoglobin polymorphisms
Age, number of measurements
aAlleles A, S and C were simultaneously analyzed.
bFBAT and QTDT were used for binary phenotype and for quantitative
cEmpirical P value for non-normal data.
Human Molecular Genetics, 2004, Vol. 13, No. 15
by guest on June 1, 2013