Prediction of Plasmodium falciparum placental infection according to the time of infection during pregnancy

Acta Tropica 98 (2006) 255–260

Prediction of Plasmodium falciparum placental infection according to the time of infection during pregnancy Gilles Cottrell a , Philippe Deloron a,∗ , Nadine Fievet a , Sokhna Sow b , Oumar Gaye c , Jean-Yves Le Hesran a a

Institut de Recherche pour le D´eveloppement (IRD), UR010, Mother and Child Health in the Tropics, Universit´e Paris Descartes, Paris, France b Thiadiaye Hospital, Senegal c Parasitology Laboratory, Cheick Anta Diop University, Dakar, Senegal Received 1 February 2006; received in revised form 18 May 2006; accepted 19 May 2006 Available online 23 June 2006

Abstract Malarial infection during pregnancy leads to placental infection, a known risk factor for low birth weight. Whether the stage of pregnancy at infection has a differential influence on these effects is not clearly known, but may be of importance for prevention strategies, including intermittent preventive treatment of pregnant women. Malaria infection during early (before 20 weeks), middle (20–28 weeks), or late (after 28 weeks) pregnancy was evaluated by logistic regression and receiver operating characteristics analysis in relation to placental infection in pregnant Senegalese women. Plasmodium falciparum infections during late pregnancy are strongly related to placental infection, as well as those that occur in middle pregnancy. Knowledge of parasitological events over the entire duration of pregnancy permits a highly accurate prediction of placental infection. Not only malaria infections during late pregnancy increase the likelihood of placental infection. The current policy of intermittent preventive treatment of pregnant women, which implies an initial antimalarial cure after 20 weeks of pregnancy, will not avoid early infections. An earlier initiation of malaria prevention might improve its efficacy. © 2006 Elsevier B.V. All rights reserved. Keywords: Malaria; Pregnancy; Placental infection; Prevention strategy; ROC analysis

1. Introduction Each year, more than 25 millions of women are exposed to malaria during pregnancy and this pathology

Abbreviations: AUC, area under the curve; IPTp, intermittent preventive treatment of pregnant women; ROC, receiver operating characteristics; SP, sulfadoxine-pyrimethamine ∗ Corresponding author. Tel.: +33 1 53 73 96 22; fax: +33 1 53 73 96 17. E-mail address: [email protected] (P. Deloron). 0001-706X/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.actatropica.2006.05.009

is estimated to cause hundreds of thousands of infant deaths (Guyatt and Snow, 2004; Murphy and Breman, 2001). The hallmark of pregnancy malaria due to Plasmodium falciparum is the accumulation of infected erythrocytes in the placenta (Garnham, 1938). Studies in various areas of stable malaria transmission reported placental infection rates at delivery varying between 10% and 34% among pregnant women of all parities (Cot and Deloron, 2003). In the recent years, the World Health Organisation implemented the intermittent preventive treatment

256

G. Cottrell et al. / Acta Tropica 98 (2006) 255–260

of pregnant women (IPTp), a new strategy to control pregnancy-associated malaria, and one of its main consequences, placental infection. This strategy relies on the administration of a curative regimen of sulfadoxinepyrimethamine (SP) after 20 weeks of pregnancy, followed by a second dose at least 1 month later. To date, only a few studies have shown that IPTp with SP was effective in decreasing the risk of peripheral parasitaemia, severe anaemia, and placental infection (Kayentao et al., 2005; Shulman et al., 1999). To better assess the efficacy of this new strategy, it is important to weigh up the relative impact of the occurrence of P. falciparum infections at various periods of pregnancy on the main indicator of IPTp efficacy, placental infection. However, P. falciparum infection during pregnancy is usually asymptomatic, and the period of occurrence of infection is difficult to identify. Few studies have investigated the relationships between P. falciparum infections during the course of pregnancy and placental infection at delivery. A study in Malawi (Wirima et al., 1993) demonstrated that they were both related. Two cohort studies, in Thailand (McGready et al., 2004) and The Gambia (Watkinson and Rushton, 1983), revealed a strong effect of malaria infection during late pregnancy (last week of pregnancy for one study and second half of pregnancy for the other) on placental parasitaemia or pigmentation. In Thailand, the last infection was recorded 1–4 months before delivery in 24% of women with an infected placenta (McGready et al., 2004). In Burkina Faso, we evaluated the proper relation between each period of pregnancy (early, middle, and late) and placental infection at delivery (Cottrell et al., 2005). Placental infection was strongly related to late malaria infection, but also independently to infection in the early pregnancy, revealing the peculiar importance of this period. Surprisingly, infection during the middle pregnancy was not significantly associated with placental infection. These relatively unexpected results needed to be confirmed by further analysis on different data sets. This is the aim of the present work, where we analyzed by logistic regression the relation between the occurrence of malaria infection during early, middle and late pregnancy and placental infection, and we quantified by receiver operating characteristics (ROC) analysis the proper importance of infection during the different periods of pregnancy in predicting placental infection. We used data from a cohort study of pregnant Senegalese women, allowing for the knowledge of parasitological events having occurred during most of pregnancy, as well as placental infection at delivery.

2. Methods 2.1. Study population and follow-up The study population and follow-up have been described elsewhere (Tuikue Ndam et al., 2006) and will be briefly summarized. We studied a cohort of 306 pregnant women in the rural area of Thiadiaye, 130 km east from Dakar, Senegal, where malaria transmission is stable and highly seasonal (from September to December). From 30 July to 15 October 2001, pregnant women presenting to Thiadiaye Hospital for an antenatal visit were enrolled if: (1) they were less than 6 months pregnant (as derived from their last menstrual period); and (2) they did not report any febrile episode since the beginning of pregnancy. As women were enrolled before or at the very beginning of malaria transmission, all presented with a negative thick blood smear and a negative rapid diagnostic test, and later confirmed by PCR. Women were followed-up until delivery during every antenatal visit and through weekly home visits, to assess the occurrence of P. falciparum infection. At each home visit, temperature was recorded. In case of fever (axillary temperature > 37.5 ◦ C), women were treated by oral chloroquine (25 mg/kg over 3 days), according to the Ministry of Health policy. A capillary blood sample was obtained for a thick blood smear, and P. falciparum PCR. In addition, passive detection of malaria attacks was implemented by asking women to present to the ANC clinic in case of any fever or sickness between the weekly home visits. At each antenatal visit, a blood sample was obtained from all women, whether they presented or not with fever. At delivery, placental infection was assessed on placental smears. The ethical committee, Ministry of Health, Senegal provided ethical approval. All participants gave informed consent. These women were not enrolled in a drug prophylaxis trial. However, according to MoH policy at time of the study, these women were told to take chloroquine prophylaxis while they were pregnant, but the drug was not given to them. Chloroquine was searched for in their urine at enrolment as well as at the following ANC, demonstrating that 45.2% and 74.8%, respectively, had taken chloroquine during the previous days. In addition, when the whole follow-up was taken into account, 92.2% of women were shown to have chloroquinuria at least once. Therefore, this variable was not taken into consideration. Similarly, no distribution of ITN has ever been done in the study area. However, each woman was asked whether she was sleeping under a bednet (impregnated or not), and when the response was yes, her home was visited to check for the presence of the bednet. A total of

G. Cottrell et al. / Acta Tropica 98 (2006) 255–260

17% of women declared to sleep under a bednet, and this bednet was present in their house. Whether they indeed slept under this bednet is unknown, but is unlikely as this variable was not related to placental infection. 2.2. Parasitological methods Malarial infection was determined on Giemsa-stained thick smears. Parasite count was estimated on 200 microscope fields, assuming a mean of 8000 leucocytes per ␮L of blood. Genomic DNA was extracted using chelex from filter paper blots, and a one-round P. falciparum PCR, using the internal P. falciparum species-specific primers FAL1 (5 -TTAAACTGGTTTGGGAAAACCAAATATATT-3 ) and FAL2 (5 -ACACAATGAACTCAATCATGACTACCCGTC-3 ) was optimized, as described (Singh et al., 1999). The PCR products were analyzed by agarose gel electrophoresis stained by ethidium bromide. 2.3. Statistical methods Placental infection (positive or negative) was studied through multivariate logistic regression as a function of peripheral blood infection with adjustment for cofactors: gestation rank (primigravidae versus multigravidae), delivery during transmission season, defined from 1 September to 31 December (yes or no), and occurrence of an attack of fever during the follow-up (yes or no). For peripheral blood infection, three binary variables (yes/no) were defined according to gestational age (determined from the date that was reported for the woman’s last menstrual period) at blood collection, less than 20 weeks (early pregnancy), 20–28 weeks (middle pregnancy), and over than 28 weeks (late pregnancy). For each given gestational period, “positive” was defined as the presence of parasites in at least one blood sample, and “negative” as the absence of parasites in all samples. In selected cases, peripheral parasitaemia could not be defined because the woman either had not entered into the study yet, or no thick blood smear was available as she had never attended any antenatal visit and did not present with fever during home visits over that period. To account for this, the “missing data indicator” method (Chavance and Manfredi, 2000) was used by adding a category for missing values (i.e. peripheral parasitaemia can take three values, negative, positive or unknown). Three multivariate logistic models were built. In each model, placental infection was regressed against different combinations of the peripheral blood infection variables, adjusted for cofactors. In model 1, variables

257

corresponding to early and middle pregnancy were introduced. In model 2, the variable corresponding to late pregnancy was introduced. In model 3, all three infection variables corresponding to early, middle and late pregnancy were introduced. This allowed us to determine the proper relation between each infection variable and placental infection, after adjustment for the two others and for cofactors. The prediction abilities of the three models were compared by testing the equality of the areas under the corresponding ROC curves, and by comparing different compromises between sensitivity and specificity for each model. A comparison of models 1 and 2 allowed us to assess the predictive power of the knowledge of early and middle pregnancy events relative to late pregnancy. Models 2 and 3 allowed us to evaluate the gain in predictive power through the addition of early and middle pregnancy information to parasitological information during late pregnancy. A non-parametric trapezoidal rule was used for estimating areas under curves (AUC). Comparisons of AUC were performed using the nonparametric method of DeLong et al. (1988). All analyses were performed using Stata 8.2 (STATA Corporation). 3. Results A total of 306 pregnant women were enrolled. Twelve were lost to follow-up. Thirteen were excluded from the analysis because the duration of pregnancy could not been determined accurately. Finally, 281 women were included for further analyses. These women were enrolled at a mean (S.D.) age of 26.4 (6.5) years, after a mean of 18.2 (4.9) weeks of pregnancy. Their pregnancy lasted for a mean of 35.8 (4.3) weeks. Their general characteristics are presented in Table 1. Models 1 and 2 are nested in model 3, which contains the complete information related to the proper effect of the three periods. Only the results of the logistic regression from model 3 are detailed. Both infection in middle and late pregnancy were related to an increased risk of placental infection (OR = 5.0 and 6.9, both p ≤ 0.001 for middle and late pregnancy, respectively). A nonsignificant OR of 1.4 was observed for early infection. Primigravid status (OR = 2.2, p = 0.06), delivery during the transmission period (OR = 4.8, p = 0.004), and occurrence of an episode of fever (OR = 2.0, p = 0.09) were also related to an increased probability of placental infection. ROC curves corresponding to each model are given in Fig. 1. Estimated AUC with 95% confidence intervals are 0.81 [0.74; 0.88], 0.82 [0.75; 0.89], and 0.86 [0.80; 0.92] for models 1, 2 and 3, respectively. The hypothesis of global equality of the three areas is rejected (p = 0.003). In further comparisons, models 1 and 2 had

258

G. Cottrell et al. / Acta Tropica 98 (2006) 255–260

Table 1 General characteristics of the women Factor Gestational rank 1 2 3–4 >4

Nb (%)

Sensitivity (%) 59 (21) 53 (18.9) 88 (31.3) 81 (28.8)

Delivery during transmission period Yes No

168 (59.8) 113 (40.2)

Febrile episode Yes No

116 (41.3) 165 (58.7)

Placental infection Yes No

45 (16.0) 236 (84.0)

Peripheral infection during pregnancy Before 20 weeks Yes No Unknown

Table 2 Specificity corresponding to different levels of sensitivity for each model

60 80 90

Specificity (%) Model 1

Model 2

Model 3

87.1 66.0 53.2

83.4 69.4 51.4

89.2 79.6 68.1

similar AUC (p = 0.76), whereas the curve from model 3 was above the other 2 (Fig. 1), with a greater AUC (model 3 versus model 2: p = 0.02, model 3 versus model 1: p = 0.03). Table 2 shows the specificity corresponding to three levels of sensitivity (chosen at 60%, 80%, 90%) given by the three models. 4. Discussion

22 (7.8) 163 (58.0) 96 (34.2)

Between 20 and 28 weeks Yes No Unknown

59 (21.0) 192 (68.3) 30 (10.7)

After 28 weeks Yes No Unknown

80 (28.4) 137 (48.8) 64 (22.8)

Fig. 1. ROC analysis of the prediction of placental infection according to the informed occurrence of P. falciparum during early and middle of pregnancy (model 1 (
)), late (model 2 (䊉)), and overall pregnancy (model 3 ()). Early pregnancy was defined as stage of pregnancy under 20 weeks, middle as stage of pregnancy between 20 and 28 weeks of pregnancy, and late pregnancy as stage of pregnancy over 28 weeks.

In this study we have aimed to determine when during pregnancy the occurrence of P. falciparum infection provides relevant information in terms of prediction of placental infection (after having adjusted for potential confounders), in order to confirm or not our recent results of another data set. An infection during late pregnancy (after 28 weeks of pregnancy) was strongly related to the presence of a placental infection, with a high predictive power (high area under the ROC curve, high specificity for a given sensitivity). Infection during the middle pregnancy was also significantly related to a higher risk of placental infection, with an almost equivalent predictive power (when associated with the knowledge of malarial status in early pregnancy). The best predictive model was given by taking into account parasitological information during whole pregnancy. Our observation that P. falciparum infection during late pregnancy is predictive of placental infection is in agreement with preceding works (McGready et al., 2004; Wirima et al., 1993). P. falciparum parasites sequester in the placenta, and it is not surprising that an infection occurring late in pregnancy is not necessarily cleared at delivery, particularly at the placenta level. Most interesting are our findings concerning middle pregnancy. Multivariate regression with adjustment for the three infection variables corresponding to the three pregnancy periods and cofactors (model 3) shows a significant and high OR for an infection in middle pregnancy. This suggests a substantial proper effect of infection during this period on placental infection, independently to infection during any other period and to confounders. This is consistent with previous works (Taha Tel et al., 1993; Brabin, 1983) suggesting that malarial infections occur-

G. Cottrell et al. / Acta Tropica 98 (2006) 255–260

ring during the last weeks of pregnancy are not alone in potentially causing harmful consequences at delivery. We have also recently shown in analysing data from Burkina Faso that women infected early in pregnancy (before the fifth month) presented with a higher risk of placental infection at delivery than women not infected during that period (Cottrell et al., 2005). In the present study, infection in middle pregnancy seems more related to placental infection than infection during early pregnancy. Because of the smallest sample size in this study and to increase the number of women with known malarial status in early pregnancy, the pregnancy periods have not been identically defined in both studies: before 16 weeks and between 16 and 26 weeks for early and middle pregnancy in data from Burkina Faso, before 20 weeks and between 20 and 28 weeks in this study. This may partially explain the differences between results of the two studies. This discrepancy between the significant periods of infection might also be related to a reduced power of the analysis of the current dataset, especially during early pregnancy. Nevertheless, both analyses are similar in demonstrating the existence of a strong effect of past infections (occurring in early or middle pregnancy) on placental infection at delivery. Both studies confirm that not only the late pregnancy, but also earlier periods, are of importance for the delivery outcome. ROC analysis allowed us to refine this result, by comparing the predictive power of infection at early and middle pregnancy (model 1) and late pregnancy (model 2). The high AUC value of model 1 shows that the knowledge of malarial status in early and middle pregnancy confers a good predictive power of placental infection, and more surprisingly, as good as the predictive power given by the knowledge of malaria events during late pregnancy (model 2). Most of this predictive power of model 1 is due to the knowledge of the malarial status in the middle pregnancy. Results from Table 1, where high sensitivity was chosen (to limit the occurrence of false negatives as much as possible) give a quantitative view of this conclusion. The specificities for given sensitivities are almost similar between both models 1 and 2. Model 3, where the three periods of pregnancy are adjusted for each other, ensures that there is a significant proper predictive power conferred by the knowledge of early and middle pregnancy, since the AUC of this model is significantly higher than in the two other models, where malarial status is only partially known. A remarkable gain of specificity is also given by model 3, as compared to both others. P. falciparum infections during middle pregnancy are associated with placental infection: this implies a long-term persistence of parasites in the placenta.

259

The pregnancy-related immune modulation, more pronounced in the placental than in the peripheral blood (Diouf et al., 2004), may facilitate this persistence in the placenta, although parasites are undetected in peripheral blood. Moreover, the presence of non-circulating, cryptic, parasite subpopulations were reported in the placenta (Pouvelle et al., 2000). Selected cofactors, primigravid status and delivery during the transmission period, were risk factors of placental infection, in agreement with numerous results (Steketee et al., 1996). Fever episodes during pregnancy were also related to placental infection, although febrile women received chloroquine, according to the policy enacted by the Ministry of Health. This certainly is likely to be a consequence of the high level of chloroquine-resistance in this area. Our two studies confirmed that P. falciparum infections during late pregnancy have consequences on placental infection. Moreover, these two studies also suggest a substantial relation between past infections (early or middle pregnancy) and placental infection, and the knowledge of malarial events occurring during these periods (especially middle pregnancy in the present study) significantly increases the predictive power of placental infection, being related to an increased morbidity and mortality of the newborn. The recommended calendar of IPTp, with the first administration of sulfadoxinepyrimethamine after 20 weeks of pregnancy, aims to avoid the period of initial development of the fetus and to target the period of maximum fetal growth. However, this strategy will not avoid early infections, and efforts should be made to consider the use of another antimalarial drug than sulfadoxine-pyrimethamine during the early pregnancy. Acknowledgements We would like to acknowledge the support of the maternity staffs and Thiadiaye health centre in the collection of samples. We are grateful to the mothers who participated in the study. We would also like to acknowledge the IRD research unit in Senegal for their practical support. This work received financial support from the Commission of the European Union (grant QLK2CT-2001-01302, PAMVAC) and the French Ministry of Research (PAL + program). References Brabin, B.J., 1983. An analysis of malaria in pregnancy in Africa. Bull. World Health Organ. 61, 1005–1016. Chavance, M., Manfredi, R., 2000. Mod´elisation d’observations incompletes. Rev Epidemiol Sante Publique 48, 389–400.

260

G. Cottrell et al. / Acta Tropica 98 (2006) 255–260

Cot, M., Deloron, P., 2003. Malaria prevention strategies. Br. Med. Bull. 67, 137–148. Cottrell, G., Mary, J.Y., Barro, D., Cot, M., 2005. Is malarial placental infection related to peripheral infection at any time of pregnancy? Am. J. Trop. Med. Hyg. 73, 1112–1118. DeLong, E.R., DeLong, D.M., Clarke-Pearson, D.L., 1988. Comparing the areas under two or more correlated receiver operating characteristic curves: a nonparametric approach. Biometrics 44, 837–845. Diouf, I., Fievet, N., Doucoure, S., Ngom, M., Gaye, A., Dumont, A., Ndao, C.T., Le Hesran, J.Y., Chaouat, G., Deloron, P., 2004. Monocyte activation and T cell inhibition in Plasmodium falciparuminfected placenta. J. Infect. Dis. 189, 2235–2242. Garnham, P., 1938. The placenta in malaria with special reference to reticulo-endothelial immunity. Trans. R. Soc. Trop. Med. Hyg. 32, 13–48. Guyatt, H.L., Snow, R.W., 2004. Impact of malaria during pregnancy on low birth weight in sub-Saharan Africa. Clin. Microbiol. Rev. 17, 760–769. Kayentao, K., Kodio, M., Newman, R.D., Maiga, H., Doumtabe, D., Ongoiba, A., Coulibaly, D., Keita, A.S., Maiga, B., Mungai, M., Parise, M.E., Doumbo, O., 2005. Comparison of intermittent preventive treatment with chemoprophylaxis for the prevention of malaria during pregnancy in Mali. J. Infect. Dis. 191, 109–116. McGready, R., Davison, B.B., Stepniewska, K., Cho, T., Shee, H., Brockman, A., Udomsangpetch, R., Looareesuwan, S., White, N.J., Meshnick, S.R., Nosten, F., 2004. The effects of Plasmodium falciparum and P. vivax infections on placental histopathology in an area of low malaria transmission. Am. J. Trop. Med. Hyg. 70, 398– 407. Murphy, S.C., Breman, 2001. Gaps in the childhood malaria burden in Africa: cerebral malaria, neurological sequelae, anemia, respiratory distress, hypoglycemia, and complications of pregnancy. Am. J. Trop. Med. Hyg. 64 (1–2 Suppl.), 57–67.

Pouvelle, B., Buffet, P.A., Lepolard, C., Scherf, A., Gysin, J., 2000. Cytoadhesion of Plasmodium falciparum ring-stage-infected erythrocytes. Nat. Med. 6, 1264–1268. Shulman, C.E., Dorman, E.K., Cutts, F., Kawuondo, K., Bulmer, J.N., Peshu, N., Marsh, K., 1999. Intermittent sulphadoxinepyrimethamine to prevent severe anaemia secondary to malaria in pregnancy: a randomised placebo-controlled trial. Lancet 353, 632–636. Singh, B., Bobogare, A., Cox-Singh, J., Snounou, G., Abdullah, M.S., Rahman, H.A., 1999. A genus- and species-specific nested polymerase chain reaction malaria detection assay for epidemiologic studies. Am. J. Trop. Med. Hyg. 60, 687–692. Steketee, R.W., Wirima, J.J., Slutsker, L., Roberts, J.M., Khoromana, C.O., Heymann, D.L., Breman, J.G., 1996. Malaria parasite infection during pregnancy and at delivery in mother, placenta, and newborn: efficacy of chloroquine and mefloquine in rural Malawi. Am. J. Trop. Med. Hyg. 55 (1 Suppl.), 24–32. Taha Tel, T., Gray, R.H., Mohamedani, A.A., 1993. Malaria and low birth weight in central Sudan. Am. J. Epidemiol. 138, 318–325. Tuikue Ndam, N., Salanti, A., Le Hesran, J.Y., Cottrell, G., Fievet, N., Turner, S., Sow, S., Dangou, J.M., Theander, T., Deloron, P., 2006. Analysis of anti-VAR2CSA IgG response in a cohort of Senegalese pregnant women. J. Infect. Dis. 193, 713–720. Watkinson, M., Rushton, D.I., 1983. Plasmodial pigmentation of placenta and outcome of pregnancy in West African mothers. Br. Med. J. (Clin. Res. Ed.) 287, 251–254. Wirima, J.J., Khoromona, C.O., Steketee, R.W., Heymann, D.L., Slutsker, L., Coker, T., O’Mahoney, G., Breman, J.G., Campbell, C.C., 1993. Malaria Prevention in Pregnancy: The Effects of Treatment and Chemoprophylaxis on Placental Malaria Infection, Low Birth Weight, and Fetal, Infant, and Child Survival. United States Agency for International Development and US Department of Health and Human Services, Atlanta, 125 pp.