Placental Malaria is Associated With Increased Risk of Nonmalaria Infection During the First 18 Months of Life in a Beninese Population
Agns Le Port
Facult des Sciences de la Sant, Universit d'Abomey-Calavi
Facult de Pharmacie,
Universit Paris Descartes
Received 27 January 2012; accepted 1 May 2012; electronically published 18 May 2012. Infections Tropicales,
Facult des sciences pharmaceutiques
, 4, Avenue de l'Observatoire, 75270 Paris Cedex 06,
Institut de Recherche pour le Dveloppement, Unit Mixte de Recherche 216
, Mre et Enfant Face aux Infections Tropicales
Centre d'Etudes et de Recherche sur le Paludisme Associ la Grossesse et l'Enfant
, Cotonou, Bnin
Background. Several studies have shown that the risk of malaria infection increases for children born to a mother with placental malaria infection. An immune tolerance phenomenon has been hypothesized. We addressed whether Plasmodium falciparum placental infection could additionally be associated with the risk of nonmalaria fevers in infants. Methods. From 2007 to 2009, 553 infants were followed up from birth to 18 months in Benin. The occurrence of fever was actively screened by trained community workers. Malaria fevers (temperature >37.5C with positive results of rapid diagnostic test or thick blood smear) were excluded from analysis. The association between placental malaria infection and the number of total, gastrointestinal, and respiratory febrile episodes was explored using binomial negative regression, with adjustment for maternal age, parity, parents' schooling, socioeconomic level, sex, village of birth, season of birth, prematurity, Apgar score and nutritional status. Results. The prevalence of placental malaria infection was 11.2%. During a median follow-up of 17.8 months, 624 nonmalaria fevers were registered. Placental malaria infection was associated with a higher risk of nonmalaria fever episodes (adjusted incidence rate ratio, 1.4; 95% confidence interval, 1.1-1.8) as well as gastrointestinal (1.6; 1.1-2.5) and respiratory (1.5; 1.1-2.1) febrile syndromes. The same pattern was obtained when considering consultations after the age of 6 months. Conclusions. These results suggest an association between placental malaria infection and nonmalaria infections in the first 18 months of life. Immune tolerance could lead to impaired immune development not specific to malaria infections in infants born to mothers with placental malaria infection, but further studies are needed. More than 240 million clinical malaria cases occurred in 2008, with approximately 1 million deaths, the vast majority occurring among young children and pregnant women in Africa . Plasmodium falciparum is the most widespread and the main species responsible for malaria's severe forms. Malaria infection during pregnancy can lead to a placental malaria infection,
which is estimated to cause up to 200 000 infant
deaths every year as a result of complications that
include stillbirth, fetal growth restriction, and preterm
delivery. The association between placental infection
and low birth weight has been well established, and
low birth weight is clearly associated with higher risk
of mortality during the first months of life .
Primigravidae are more susceptible to placental
P. falciparum infection, and they and their offspring
are therefore proportionately more affected by these
adverse outcomes than are multigravidae .
To date, 4 studies have shown that children born of a
mother with an infected placenta seem to have an
increased risk of malaria infection , and it has been
conjectured that placental infection may alter infants
immunological response and be responsible for an
immune tolerance phenomenon . However, although the
ability to construct a P. falciparumspecific immune response is
acquired by the fetus very early in utero, to our knowledge, no
clear hypothesis has been put forward to clarify the underlying
mechanisms that could lead to immune tolerance . The
association between placental malaria infection and immune
development has already been described [13, 14] as well as the fact
that malaria infection can suppress immunity to a variety of
viral infections, such as human immunodeficiency virus,
Epstein-Barr virus, vaccinia virus, and lymphocytic
choriomeningitis virus infections . More recently, it has been
shown that congenital cytomegalovirus infection was more
frequent in children born to mothers with P. falciparum
infected placenta . In this last study, this association was
adjusted for maternal age, bed net use and other environmental
risk factors; moreover, during the first year of life, mothers of
congenitally infected children reported more health complaints
for their child. Altogether, these results strengthen the potential
implication of a tolerogenic environment caused by malaria
placental infection in this phenomenon.
Therefore, the question to be answered is whether this
tolerogenic environment could be responsible for an increased
susceptibility to malaria only or, more generally, to infections.
This study explores the potential effect of P. falciparum
placental infection on the occurrence of nonmalaria fevers in
infants from birth until 18 months of life.
Study Area and Population
This study, forming a part of a larger project , was
conducted in 9 villages of the Tori Bossito area (Southern Benin)
and 3 health centers (Tori Avame, Tori Cada, and Tori Gare)
providing birth attendance and primary healthcare. All
women living in any of the 9 villages and attending health
centers for prenatal care were recommended to participate in
the study with their offspring after delivery. The birth cohort
was initiated in June 2007, and subjects were recruited until
July 2008. Figure 1 shows the flow chart of the study
population. Of the 656 pregnancies registered, 26 multiple
pregnancies and 23 stillbirths or neonatal deaths (<28 days) were
excluded, as were children with identification number errors
(n = 13) or missing birth date (n = 3), leaving 564 infants.
Among those, maternal placental infection was missing in
11 (2%); these subjects were not considered in this analysis
either. Finally, 553 infants were studied.
end of this questionnaire, and before delivery, anthropometric
measurements (weight and height) were taken. After delivery,
thick and thin blood smears were performed from the maternal
side of the placenta to determine the existence of a malaria
placental infection. Gestational age at delivery was determined using
Ballards method . Newborns were listed and received an
identification card, giving them access to free treatment in health
centers during the 18 months of follow-up.
Each child was visited once a week and axillary temperature
was measured with a digital thermometer by community
health workers at the infants home (active fever screening). In
case of temperature higher than 37.5C, mothers were told to
bring their children to the health center where a questionnaire
was filled out. Mothers were also invited to bring their infants
to the health center at any time for free care in case of fever
(suspected by the mother) or clinical signs. To test for malaria
infection, both a rapid diagnostic test (RDT) and a thick
blood smear (TBS) were performed; the TBS was stained with
Giemsa stain and read by 2 laboratory technicians (with <1%
disagreement). Leukocytes and parasites were counted
simultaneously. A TBS was declared negative if no parasite
was found after 500 leukocytes had been counted.
Symptomatic malaria infection was defined as the presence of fever
(>37.5C) and positive findings at either TBS and/or RDT. In
case of symptomatic malaria infection, infants were treated
with an artemisinin-based combination therapy (artemether
and lumefantrine) as recommended by the Benin National
Malaria Control Program. All health expenditures, whatever
the cause, were assumed by the program.
At delivery, a questionnaire was administered to gather
information on womens sociodemographic, gynecological, and obstetric
background and on the course of the current pregnancy. At the
Definition of the Outcomes
The outcome of interest was the nonmalaria fever episodes
occurring during the first 18 months of life. They were defined
by the presence of a measured axillary temperature 37.5C,
negative results of malaria RDT, and negative TBS during the
consultation. When only the RDT or the TBS result was
available, we considered its result.
In the second step, 2 types of nonmalaria febrile syndromes
were considered: gastrointestinal and respiratory. The first was
defined as the association of nonmalaria fever and 1
gastrointestinal symptom (vomiting, diarrhea, constipation or
abdominal pain). The second was defined as the association of
nonmalaria fever and 1 respiratory (cough, dyspnea or
abnormalities on auscultation) or 1 otorhinolaryngological clinical
sign (mainly otitis or rhinitis). One given febrile episode could
match both definitions of gastrointestinal and respiratory
syndromes; in this case it was included in both categories.
Explicative variables concerning the mother, the family and
the child were collected at inclusion and during follow-up.
Mother and Family Variables. The following variables
were collected: (1) age of the mother at delivery (in years);
(2) parity ( primiparae vs multiparae); (3) educational status of
the parents (1 parent schooled vs neither parent schooled);
(4) and socioeconomic level evaluated according to the
number of durable goods (radio, motorbike, flashlight, bicycle,
car, pirogue, electric generator, etc) owned by the family
estimated on a declaration basis (<4 or 4 goods).
Child Variables. Concerning the child, we collected the
following variables. (1) sex, (2) village of birth, (3) season of
birth (during rainy season or not), (4) prematurity (defined as
a gestational age <37 weeks), (5) the 5-minute Apgar score
(7 vs >7), and (6) nutritional status based on World Health
Organization (WHO)/UNICEF indicators (characteristics of
breast-feeding and infant and young child feeding) . These
indicators are based on the practice of exclusive breast-feeding
from birth to 6 months and on the number of food groups
consumed up to 18 months (assessed by individual 24-hour
dietary recall of the mother) as recommended by WHO.
As the dependent variables are nonnegative counts, we first
used Poisson regression models. However, goodness-of-fit
tests revealed that the dependent variables were not consistent
with a Poisson-like process because of an overdispersion (data
not shown). We then used negative binomial regression
models, which allowed us to take this overdispersion into
account. Results were expressed as incidence rate ratios (IRRs)
rather than the underlying coefficients, and the potential
variability of the duration of follow-up for each individual was
taken into account.
The association between P. falciparum placental infection
and the number of nonmalaria febrile episodes was analyzed
before and after adjustment for measured covariates, globally
and for each syndrome.
We then re-ran the analysis without considering the events
that occurred during the first 6 months of life to take into
account the potential effect of maternal immunoglobulin that
could confer a passive immunity to the child. All these analyses
were performed with Stata software, version 11.0 (StatCorp).
The study protocol was approved by the Ethics Committee of
the University of Abomey-Calavi (Facult des Sciences de la
Sant) in Benin and the Consultative Committee of Ethics of
Institute of Development Research.
Description of the Population
Among the 4602 consultations that occurred during the
follow-up, 711 were follow-up visits scheduled by nurses to
evaluate the childs health following a previous consultation.
They were excluded from the analysis. Among the remaining
3891 consultations, fever was recorded in 1212, including
624 with a negative malaria RDT and TBS results. The
median follow-up duration was 17.8 months (interquartile
range, 17.418.1). Placental malaria infection had no
significant effect on this duration.
The prevalence of placental malaria infection was 11.2% (62
of 553 infants). Placental malaria infection was significantly
associated with a younger maternal age, primiparity, and birth
during the rainy season (Table 1). The associations between
placental malaria infection and socioeconomic level and
village of birth were borderline nonsignificant (P = .060 and
P = .057, respectively).
In addition, 63.1% of the infants had 1 nonmalaria fever
episode during follow-up. Among them, the median number
of nonmalaria fever episodes was 1 (interquartile range, 12).
The maximum numbers of total, gastrointestinal and
respiratory febrile episodes were 6, 3, and 5, respectively. Incidence
rates of febrile episodes in the study population are plotted
in Figure 2. The incidence rate was higher for children born
from a mother with an infected placenta during the entire
Risk Factors of Nonmalaria Fever Episodes
Multivariate analyses showed that placental malaria infection
was significantly associated with a higher risk of nonmalaria
fever (P = .019). The village of residence was strongly
associated with fever episodes, consistent with the existence of a
spatial heterogeneity of risk (P = .0008). There was a
nonsignificant trend in favor of an association between prematurity
and nonmalaria fever (P = .066).
Table 1. Characteristics of Infants According to Plasmodium
falciparum Placental Infection
P. falciparum Placental
Characteristic Yes (n = 62) No (n = 491) Maternal age, years
Data represent No. (%) unless otherwise specified.
a P values are based on (or Fisher exact) tests for qualitative data and
Mann-Whitney tests for quantitative data.
b WHO/UNICEF Infant and young child feeding score.
The same pattern of association was observed between
placental malaria infection and gastrointestinal or respiratory
febrile syndrome (P = .020 and P = .006, respectively).
Concerning gastrointestinal syndrome, the only other factor significantly
associated with nonmalaria fever was the childs 5-minure
Apgar score (P = .007). Finally for respiratory febrile episodes,
Figure 2. Incidence rates of febrile episodes according to placental
malaria infection. Quarterly incidence rates for subjects born to mother
with (white circles) or without (black squares) placental Plasmodium
apart from placental malaria infection, prematurity was the
only significant association (P = .023), despite a nonsignificant
trend in favor of an association with sex (P = .073) and the
childs 5-min Apgar score (P = .081).
As shown in Table 2, the incidence of febrile episodes was
9.03 per 100 person-months in case of placental malaria
infection and 6.83 per 100 person-months in the population
without placental malaria infection. The infants born to
mothers whose placenta was infected by P. falciparum
presented 1.3-fold more febrile episodes than the others (IRR,
1.3, 95% confidence interval [CI], 1.01.7). This association
persisted after adjustment (IRR, 1.4, 95% CI, 1.11.8).
Moreover, when gastrointestinal and respiratory syndromes were
considered, the same significant associations were observed
(adjusted IRR, 1.6, [95% CI, 1.12.5] for gastrointestinal and
1.5 [1.12.1] for respiratory syndromes). The goodness of fit
of the 3 models, represented in Figure 3, is consistent with
the fact that the binomial negative model was well suited to
The same pattern of results was obtained when considering
only consultations after the age of 6 months (data not shown).
Our main result was that placental malaria infection is
strongly associated with an increased risk of appearance of
nonmalaria fever episodes between birth and 18 months of
life. This association remained significant after 6 months
of age, suggesting that the effects cannot be explained solely
by the immunity potentially induced by maternal
immunoglobulins during the first months of life. Furthermore, this
association was still significant despite a strong influence of the
Table 2. Incidence Rates of Nonmalaria Febrile Episodes and
Incidence Rate Ratios of the Plasmodium falciparum Placental
P. falciparum Placental Infection
Yes (n = 62)
No (n = 491)
Abbreviations: CI, confidence interval; IRR, incidence rate ratio.
a Incidence rates globally estimated on the sample.
b IRR from binomial negative regression model adjusted for age of the
mother at delivery, parity, parents schooling, socioeconomic level, sex,
village of birth, season of birth, prematurity, 5-minute Apgar score, and
village of residence in multivariate analyses, also suggesting a
spatial variation of childhood fever morbidity in the area
previously reported in Malawi and Nigeria . There was
no significant interaction between gravidity and placental
infection on the occurrence of nonmalaria fever (P = .64). When
we considered fever episodes as related to gastrointestinal or
respiratory clinical signs, similar associations were obtained
with placental malaria infection. Interestingly, among the
other risk factors evaluated in this study, prematurity and
Apgar score, both related to the childs potential frailty,
presented a significant association with nonmalaria fever
Our protocol was particularly well suited to detecting
nonmalaria fevers in the study population during the entire
follow-up period. Indeed, not only was the recruitment
quasiexhaustive within the study area and during the inclusion
phase, but fever episodes were also actively screened once a
week by means of a health agent network, and mothers were
invited to bring their infants to the health center for free
attendance in case of suspected fever. Furthermore, every day,
our team checked that the children in whom a fever had been
diagnosed the day before consulted at the dispensary. If not,
our team went to the family to evaluate the health status of
the child and to perform RDT and TBS for malaria diagnosis.
This protocol, different from the majority of programs
focusing on the determinants of fever, leads us to believe that the
vast majority of fevers were detected.
The protocols classically used are house-to-house surveys,
based on interviews with parents, or community surveys in
healthcare structures, both likely to generate bias [25, 26],
with the objective of supplying information to health policy
makers . We found no influence of socioeconomic and
educational factors with the occurrence of nonmalaria fever,
consistent with most studies in developing countries .
This could reflect the narrow gap in the socioeconomic
categories in the households of the study areas. A recent study
showed that the primary reason given for not visiting the
hospital when ill was the distance to the facility from
home [32, 33]. We did not find this association, which could
be the result of the frequent visits the team made to childrens
homes. This very close follow-up together with the fact that all
care was free could be responsible for a possible
overestimation of the number of consultations. Nevertheless, since
neither the women nor the health teams knew whether or not
the placenta was infected, the extent of the potential
overestimation of consultations was probably similar in the 2 groups
(ie, those with and those without placental malaria infection).
Because the study had been designed primarily for malaria
risk factor analysis, otorhinolaryngological and
bronchopulmonary infections could not be differentiated with sufficient
accuracy, since for instance, respiratory rate was not collected
and thoracic radiography was not available . Likewise,
both gastrointestinal and respiratory syndromes could
simultaneously occur at a given consultation. The definition of the
syndromes was therefore unspecific and the 3 outcomes were
not independent. However, taking this shortcoming into
account, we are fully convinced that the broader definition of
nonmalaria fever (ie, axillary temperature >37.5C with
negative results of malaria RDT and TBS), the main outcome of
the study, can be supported by our data. For this reason, we
consider that the main objective has been met and that
placental malaria infection is positively associated with the
occurrence of nonmalaria fever during the first months of life.
The role of differential exposure to infectious diseases
cannot be entirely removed. However, as shown by our
multivariate analysis, the village of residence and placental infection
were associated with the risk of nonmalaria fever. This result
is consistent with the existence of both a potential spatial
variation of exposure and the role of placental infection.
To our knowledge, this is the first time that such an
association has been clearly shown. Inversely, the association
between placental malaria infection and a higher susceptibility
to malaria has been described several times, including in this
population . It has been conjectured that placental
infection may alter infants immunological response and be
responsible for an immune tolerance phenomenon . In the
present study, we suggest that immune tolerance could be
responsible for the development of a tolerogenic environment,
involving a number of immune effectors  that could
interact with the development of immunity during the first
years of life. Such interactions can concern not only in the
malaria-specific immune response but also in general
immunity. For example, it has been shown that placental malaria
infection was positively associated with a higher risk of
congenital cytomegalovirus infection in West Africa . Our
results strongly support this hypothesis and need to be
confirmed, because the same pattern of results could be explained
by differences in individual susceptibility to infections
independent of placental malaria, which could be mediated by
host genetic polymorphism, as already shown for malaria .
These results suggest an association between P. falciparum
placental infection and nonmalaria infectious morbidity in the
first 18 months of life. It could be related to immune
tolerance, leading to impaired immune development that is
not specific to malaria infections in infants born to mothers
with infected placentas.
Acknowledgments. We thank the entire staff of the 3 maternity
hospitals (Avam, Cada, Gare) and the study population. We gratefully thank
our field team, especially Edgard Godonou, Stphane Gehou, Sylvestre
Zehounkpe, Pierre Adanchoedo, and Patrick Pomalegni, and the Centre
National Hospitalo Universitaire of Cotonou, who followed up and cared
for the cohort.
Financial support. This work was supported by the Agence Nationale
de la Recherche (SEST 2006/040/00), Ministre des Affaires Etrangres
(France; project REFS No. 2006-22), and the Institut de Recherche pour le
Developpement, which contributed to the study financially and with
Potential conflicts of interest. All authors: No reported conflicts.
All authors have submitted the ICMJE Form for Disclosure of Potential
Conflicts of Interest. Conflicts that the editors consider relevant to the
content of the manuscript have been disclosed.