Plasmodium vivax and Plasmodium falciparum infections in the Republic of Djibouti: evaluation of their prevalence and potential determinants

The prevalence rate of P. falciparum and P. vivax infections among adults aged 15–54 years living in the Republic of Djibouti was estimated using a anonymous non-correlated cluster sampling method between the 24^th and the 31^th of March 2002. In brief, 30 clusters were investigated in the city of Djibouti, and 25 clusters in the other districts of the country. The clusters were randomly selected proportionally to the population size according to the list of quarters used by the National Direction of Statistics in the city of Djibouti and the list of the towns used by the expanded programme of immunization in the other districts. In each selected site, a starting household was randomly selected, and the next nearest households were investigated until a total of 44 resident adults per cluster in the city of Djibouti or 35 resident adults per cluster in the other districts were obtained. A total of 1,910 blood samples were collected anonymously in accordance with the recommendations of the Djiboutian Ministry of Health, who gave the ethics clearance for the present study. Blood samples were stored at 4 °C until separation of plasma by centrifugation (less than 24 hours after collection) and freezing. Thirty sera samples from French adults who had never been to malaria-endemic countries were used as unexposed negative controls. For the seropositivity threshold, the means and standard deviations (SDs) of the antibody intensity of the negative control group for all antigens were estimated. The lower limit of positivity for each antibody was taken as the mean + 3 SD of the negative control group values and corresponded to a mean fluorescence intensity (MFI) of 1,000. For P. falciparum, a sample was considered to be positive if the reactivity to at least two different plasmodial antigens was > 1,000 MFI. For P. vivax, a sample was considered to be positive if the reactivity to PvMSP1-19 or to PvMSP1-42 was > 1,000 MFI.

Eleven Peptides (Lsa1-41, Lsa1-J, Lsa3-NR2, Glurp, GlurpP3, Salsa1, Salsa2, Trap1, Starp-R, CS (NANP) and SR11.1) were synthesized with an added N-terminal cysteine residue and covalently coupled with BSA (bovine serum albumin, Sigma-Aldrich, St. Louis, USA) by Genpep (Ales, France) and stored in aliquots at −20 °C. P. falciparum merozoite surface protein 1–19 (MSP1₁₉) and apical membrane antigen 1 (AMA1) were obtained using procedures previously described[14]. For P. vivax, MSP1₁₉ proteins were produced in a baculovirus[21] and MSP1₄₂ as described elsewhere[19].

Peptides and proteins were coupled to beads as described by Ambrosino et al.[22], and an optimal concentration of 0.3 nmol was used for each antigen. Furthermore, BSA coated beads were included as a background control. Ag-coated beads were resuspended by vortexing and sonication for 5 minutes and were diluted in equal volumes of PBS and MFIA (Multiplexed Fluorescence ImmunoAssay) diluent (Charles River Laboratories Inc, MA, USA) to a final concentration of 80 beads/μl per peptide. The 1.2-μm filter-bottom 96-well microtiterplates (MSBVS 1210, Millipore, MA, USA) were rewetted with washing buffer (0.15% Tween 20 in PBS 7.4) using a vacuum manifold (Millipore). Fifty microliters of beads and sera (diluted 1:100 in equal volumes of PBS and MFIA diluents) were added to each well. Plates were incubated at room temperature in the dark for 1 h with shaking at 600 rpm. After incubation, plates were washed eight times with 200 μl of washing buffer, then 100 μl of the secondary Ab (R-phycoerythrin F(ab’)2 fragment of goat anti-human IgG, (Interchim, Montluçon, France), diluted 1:500, was added to each well. After 30 min of incubation in the dark at room temperature with shaking, plates were washed as described previously. Beads were resuspended in 100 μl of a solution of 5% BSA-PBS, pH 7.4 and analysed on Luminex system. The system was set to read a minimum of 100 beads per spectral address, and the results were expressed as MFI.

During the cross-sectional study between the 24^th and the 31^th March 2002, self-administered questionnaires containing several items were filled out by Djiboutian inhabitants and validated by a member of the research team. Different types of independent variables were collected: the living area (the city of Djibouti or the rest of the country), the type of living area (urban or rural), having stayed in a malaria-endemic country for more than one year (yes or no), having had a fever during the last month before the study (yes or no), the utilization of bed nets (often to always and rarely to never), gender (male or female), schooling status (schooled or never schooled), educational level (never schooled, primary school, secondary school, high school or university), wealth (poor = less than 65,000 Djiboutian Francs per home and rich = more than 65,000 Djiboutian Francs per home), and age.

A Geographic Information System was built in ArcGIS 9.2 (Environmental Research Systems Institute, Redlands, CA). The layers were added as follows: i) map of inland water in Djibouti and the neighbouring countries (data from Digital Chart of the World, accessed through DIVA-GIS[23] and ii) 60 points corresponding to the sampling locations. At every point, the Euclidian distance to the first pixel of water was computed. This enabled the creation of a geographical independent variable, the distance of a cluster to a river or a lake (≤ 1.5 Km or > 1.5 Km).

Data were recorded using Excel and were checked for consistency before statistical analysis using R software (version 2.10.1) or STATA software (version 11). The seropositivity to P. falciparum antigens, P. vivax antigens or both of them (mixed infection) were analysed as a dependant variable according to individual and cluster characteristics using a random effect mixed logistic regression model. The model was designed to take into account the intracluster correlations that could exist due to the sampling design (cluster effect as random effect). The logistic model was also adjusted using a generalized estimating equations (GEE) approach. Random effect and GEE regression models allow the estimation of cluster-specific and population-averaged effects, respectively[24]. First, a descriptive analysis of the independent variables was performed. A bivariate analysis was then conducted by entering each independent variable in a logistic regression model, and all the results were presented in Additional files1,2,3,4,5 and6. Variables were retained for the multivariate analysis if their effect had a p-value less than 0.25[25]. A backward stepwise selection procedure was applied to retain the significant (p < 0.05) independent variables and their interactions in the final model. The statistical quality of the final model was assessed by looking at the adequacy between observed and predicted prevalence rates.

As the immune response to several P. falciparum or P. vivax antigens could be quantified by MFI, another ordinal dependant variable in four classes was created: the level of immune response to P. falciparum antigens (L0 = seronegativity MFI < 1000, L1 = 1000 ≤ MFI < 3800, L2 = 3800 ≤ MFI < 8000, and L3 = MFI ≥ 8000) and P. vivax antigens (L0 = seronegativity MFI < 1000, L1 = 1000 ≤ MFI < 2000, L2 = 2000 ≤ MFI < 10000 and L3 = MFI ≥ 10000). The same type of analysis was applied to this dependent variable as described above for the seropositivity status, using the “svy” command of STATA (i.e., using the linearized variance estimator based on a first-order Taylor series linear approximation) to take into account the cluster effect.

The results of the bivariate Bayesian ordinal multinomial regression analysis are presented in Additional files7 and8.

The serological analysis showed that 25.90% of sera were positive for at least two of the following 11 P. falciparum peptides: Lsa1-41, Lsa1-J, Lsa3-NR2, Glurp, GlurpP3, Salsa1, Salsa2, Trap1, Starp-R, CS (NANP) and SR11.1. The proportion of sera that were positive for both and at least one of the two recombinant falciparum antigens, i.e., PfMSP1 and PfAMA1, were 13.24% and 29.98%, respectively. By taking into account the immune humeral response to at least two different peptides or recombinant proteins of the 13 P. falciparum antigens used in this study, the seropositivity rate to P. falciparum infection was 31.5% (602/1910 Djiboutian people), CI95% (29.4-33.7).

Considering the potential determinants of the P. falciparum malaria infection, according to the results of the multivariate logistic regression analysis (Table1), some factors were significantly and independently statistically associated with a lower risk of being seropositive for P. falciparum:

In contrast, only one risk factor for seropositivity for P. falciparum was identified: older age (40 years to 55 years). There was a non-significant association of living near lakes and rivers (≤ 1.5 Km) with a higher risk of seropositivity.

By taking into account the reactivity against PvMSP1₄₂ or PvMSP1₁₉, the global seropositivity rate to P. vivax was 17.5% (334/1910 Djiboutian people), CI95% (15.8-19.3). Considering the potential determinants of the P. vivax malaria infection, according to the results of the multivariate logistic regression analysis (Table2), some factors were significantly and independently statistically associated with a lower risk of being seropositive for P. vivax:

Having no schooling was significantly and independently statistically associated with a higher risk of being seropositive for P. vivax infection.

The global seropositivity rate of mixed infection was 10.2% (195/1910 Djiboutian people), CI95% (8.9-11.7).The results of the multivariate logistic regression analysis for predicting a mixed infection (Table3), i.e., seropositivity to serological markers of both P. falciparum and P. vivax, showed that some variables were significantly and independently statistically associated with a lower risk of being seropositive for mixed infection:

Some factors were significantly and independently statistically associated with a higher risk of being seropositive for mixed infection:

The proportions of P. falciparum seropositives to the different peptides and recombinant proteins from the different L-groups (L0 = seronegativity MFI < 1,000, L1 = 1,000 ≤ MFI < 3,800, L2 = 3,800 ≤ MFI < 8,000, and L3 = MFI ≥ 8,000) were illustrated in Figure1. The distribution of corrected MFI values against the P. falciparum antigens in the L1, L2 and L3 groups was presented in Figure2.

According to the results of the Bayesian ordinal multinomial logistic regression analysis (Table4), some factors were significantly and independently statistically associated with a lower risk of having a high level of immune humeral response to P. falciparum antigens:

In contrast, only one risk factor was significantly and independently associated statistically with a risk of having a high level of immune humeral response to P. falciparum antigens:

The proportions of P. vivax seropositives to the different recombinant proteins from the different L-groups (L0 = seronegativity MFI < 1,000, L1 = 1,000 ≤ MFI < 2,000, L2 = 2,000 ≤ MFI < 10,000 and L3 = MFI ≥ 10,000) were illustrated in Figure1. The distribution of corrected MFI values against the P. vivax antigens in the L1, L2 and L3 groups was presented in Figure3. According to the results of the Bayesian ordinal multinomial logistic regression analysis (Table5), some factors were significantly and independently statistically associated with a lower risk of having a high level of immune humeral response to P. vivax antigens:

Two risk factors were significantly and independently statistically associated with a risk of having a high level of immune humeral response to P. vivax antigens:

The different clusters of P. falciparum and P. vivax seroprevalence were presented in Figures4,5,6, and7 and in Additional file9. For P. falciparum, the city of Djibouti showed a clustering of low and medium prevalence areas on both sides of the Ambouli wadi. A hotspot was observed in Arhiba, in which more than half of the population (56.4%) (Additional file10) was seropositive for P. falciparum. A mean tendency was observed in the upper town (i.e., Quarters 1 to 15), which globally showed a decreasing prevalence when the distance to Ambouli wadi increased. This tendency was also observed in the lower town, i.e., the other side of the Ambouli wadi quarter, with 4 hotspots (Balbala 2, PK12, Balbala 3, and North of Wahle Daba) that had similar prevalence to Arhiba. The most prevalent cluster in the entire country was Balbala 4, in which almost two in three persons (67.5%) were seropositive for P. falciparum.

In the rest of the country (Additional file11), the 2 regions in the north (Tadjourah and Obock) exhibited a low prevalence, except for Balho, in which more than one in three persons were seropositive for P. falciparum. The situation in the South was more concerning, as 3 seropositivity hotspots were observed in Dagguirou (46.9%), Tammiro (64.7%) and As-Eyla (64.7%) in the Region of Dikhil, and 2 seropositivity hotspots were observed in Ali-Sabieh1 (52%) and Ali-Sabieh 3 (45.5%), the capital city of the Region of Ali-Sabieh.

Considering the prevalence of P. vivax, in Djibouti city, on both sides of Ambouli wadi, an increase in the distance between quarters and the wadi was associated with a decrease in prevalence rates, as was true for P. falciparum. In the upper town, the only hotspot was in Arhiba, with a 33.3% seroprevalence. In the lower town, two hotspots were observed in Balbala 4 and north of the Wahle Daba, with 35% and 44.4% seroprevalence, respectively.

Of the northern regions, Obock and Tadjourah, the seroprevalence rates in some localities were 20% to 30%, such as in Malâho, La’Assa (Obock Region) and the regional capital Tadjourah (Tadjourah Region). In southern regions, the seroprevalence rates were between 20% and 30%, such as in Dagguirou, Garabayis (or Gour’abouss) and As-Eyla for the Dikhil region and Doudoub Balaleh for the Ali-Sabieh region. Hotspots were observed in Tammiro, with a 32.4% seroprevalence rate (in the region of Dikhil), and in Ali-Sabieh 1, with a 40% seroprevalence rate (the capital of Ali-Sabieh’s region).

The use of different antigenic peptides was dictated by the fact that in countries where malaria transmission occurs, the results of serology may be ambiguous due to cross reactions with other parasitic infections, such as toxoplasmosis[26]. Thus, by increasing the number of antigens and considering the sera reaction to at least two different plasmodial antigens, this enabled limiting the false positive rates. However, this approach using solely antigenic peptides could lead to an underestimation of the level of transmission; thus, other serologic markers were included, such as the recombinant proteins MSP1 and AMA1, the use of which there is a consensus in the literature[27‐29].

In 2009, Noor et al. observed a P. falciparum seroprevalence rate of 14.2% in adults above 50 years, 6.9% in children and an average of 9.9% when they tested the reactivity of 4769 sera to one or both serological markers PfMSP1 and PfAMA1 among Djiboutian population[14]. The same method was applied to the present work and produced a seroprevalence rate of 30.0%. Moreover, when the reaction to at least one marker of the 11 peptides, PfMSP1 and PfAMA1 were combined, the seroprevalence increased to 56.6%; finally, a seroprevalence rate of 33.5% was obtained when considering reactions to at least 2 of the 13 markers. In light of these observations, it can be deduced that the prevalence and, indirectly, the P. falciparum malaria transmission have declined by at least a factor of three in the past seven years. These results were consistent with the needs and obligations that lead to a pre-elimination goal in which the reduction of transmission is the most important key to pre-elimination[15].

All obtained models predicting the P. falciparum or P. vivax seropositivity status or the level of humeral immune response to P. falciparum or P. vivax antigens have shown that female gender, a high educational level and never having visited a malaria-endemic country more than one year were protective. Considering the mixed infection seropositivity status, the multivariate logistic regression model showed a protective effect of living at a distance > 1.5 Km from rivers and lakes, in a rural area and not having had fever during the last month before the cross-sectional study. Only the model that predicted the P. falciparum seropositivity status showed a protective effect of younger ages between 15 and 40 years. As serological tools reflect the cumulative exposition[28], these observations suggest that the older populations were more exposed and that transmission was thus higher in the past. In the city of Djibouti, educational level is generally correlated with the level of wealth and therefore more accessibility to health facilities and prevention measures[30]. Historically, Ethiopia and Djibouti have maintained very important population exchanges in both directions. Therefore, it is normal to see certain Djiboutian populations settle there for long periods due to the far lower living costs when facing economic or social difficulties in the Republic of Djibouti[31]. Carteron in 1978 and Fox in 1991 have shown that Ethiopia was the most important provider of malaria cases to Djibouti[7, 32]. This may explain the observation that living in malaria-endemic country (and especially Ethiopia) for more than one year was a risk factor for being seropositive to P falciparum or P. vivax.

In the city of Djibouti, the seroprevalence rates for both parasite species revealed hotspots on both sides of the main wadi, i.e., Ambouli wadi and the quarters of Arhiba and Balbala 4. Arhiba and Balbala 4 are quarters with significant migrant populations who regularly travel to and/or from Ethiopia[31]. There was an association between the decreasing of seroprevalent clusters to both species and the increase of the cluster distance to Ambouli wadi.

In the rest of the country, P. falciparum seroprevalence rates were higher in the southern regions (Dikhil and Ali-Sabieh), and in particular, hotspots were found along the land routes to Ethiopia, i.e., Tammiro/As-Eyla and Ali-Sabieh. These roads are regularly used by professional truckers, private users and migrants because they are the only two terrestrial roads to Ethiopia[33].

Plasmodium vivax seropositivity status was more balanced throughout the Djiboutian territory, with hotspots in the same locations in the southern regions as for P. falciparum. Because of the possibility of liver persistence of hypnozoïtes, P. vivax can spread more widely across the entire country in the case of incomplete treatment. P. vivax can be found where P. falciparum is no longer detectable and can sometimes be more prevalent, as could be the case in the neighbouring countries[34]. This situation might explain the high seroprevalence rates recorded in northern regions and suggests the occurrence of local transmission foci when Anopheles vectors exist.

Pre-elimination is a combination of technical feasibility, i.e., the baseline domestic malaria transmission combined with the importation-related transmission, and operational feasibility, which takes into consideration the country’s government status, health status and information on populations at risk[13].

Compared to one recent work[14], the present serological study indirectly indicates that transmission may have decreased by three-fold in the past seven years for P. falciparum, even though some hotspots were the same as those found in the Dikhil region. This work, once combined with recent information provided by Noor et al. has shown that the Republic of Djibouti is likely on the correct path to pre-elimination with benefits that are threatened by the persistence of hotspots such as those in the Dikhil Region. Finding the same hotspots seven years later constitutes a serious threat to the success of the announced goal. Taking into account technical and operational feasibility, pre-elimination is possible only insofar as efficient control methods are implemented at all administrative and executive levels of authority.

Recommendations emerge from the results of the statistical models. If the educational level of the population cannot be increased, pre-elimination will require increased awareness and health education for at-risk populations through methods that are appropriate to local realities. Although being female seems to be protective, health education should target mothers in particular to increase local knowledge, as has been previously done with HIV[35]. Regional collaborative alert systems are also indispensable because the population is regularly moving to a neighbouring country. To act efficiently on hotspots, the mass distribution of bed nets, mass chemotherapy and chemoprophylaxis for both parasites and indoor residual spraying should be performed, followed by serosurveys and vector monitoring. The vector monitoring system is actually debutant and in progress, although it primarily concerns Djibouti city. Border posts should also see an improvement of their monitoring activities and health controls.