Global trend of Plasmodium malariae and Plasmodium ovale spp. malaria infections in the last two decades (2000–2020): a systematic review and meta-analysis

Recent studies indicate that the prevalence of non-falciparum malaria, including Plasmodium malariae and Plasmodium ovale spp., is increasing, with some complications in infected individuals. The aim of this review is to provide a better understanding of the malaria prevalence and disease burden due to P. malariae and P. ovale spp. The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines and the Joanna Briggs Institute prevalence study assessment tool were used to select and evaluate the studies, respectively. Six databases: PubMed, WHOLIS, Wiley Library, ScienceDirect, Web of Science and Google Scholar were used to screen articles published during the period January 2000–December 2020. The pooled prevalence estimates for P. malariae and P. ovale spp. were analysed using a random-effects model and the possible sources of heterogeneity were evaluated through subgroup analysis and meta-regression. Out of the 3297 studies screened, only 113 studies were included; among which 51.33% were from the African Region. The P. malariae and P. ovale spp. pooled prevalence were 2.01% (95% CI 1.31–2.85%) and 0.77% (95% CI 0.50–1.10%) respectively, with the highest prevalence in the African Region. P. malariae was equally distributed among adults (2.13%), children (2.90%) and pregnant women (2.77%) (p = 0.862), whereas P. ovale spp. was more prevalent in pregnant women (2.90%) than in children ≤ 15 years (0.97%) and in patients > 15 years old (0.39%) (p = 0.021). In this review, data analysis revealed that P. malariae and P. ovale spp. have decreased in the last 20 years, but not significantly, and these species were more commonly present with other Plasmodium species as co-infections. No difference in prevalence between symptomatic and asymptomatic patients was observed for either P. malariae or P. ovale spp. Our analysis suggests that knowledge of the worldwide burden of P. malariae and P. ovale spp. is very important for malaria elimination programmes and a particular focus towards improved tools for monitoring transmission for these non-falciparum species should be stressed upon to deal with increased infections in the future.

in African and South-East Asia regions [1]. Several international programmes have been launched to reduce and/ or eradicate malaria worldwide, and considerable progress has been made over the past two decades, resulting in a decrease in the global incidence of malaria from 363 to 225 cases per 1000 population during the years 2000 to 2019 [1]. Unfortunately, recent reports of the emergence of artemisinin resistance in South-East Asia in the Greater Mekong Subregion adds another challenge hindering the efforts of malaria control programmes [2][3][4].
Among all human Plasmodium spp., P. falciparum and P. vivax receive particular scientific research interest, as they are responsible for the majority of malaria infections and disease severity in human patients, whereas P. ovale spp. and P. malariae remain neglected tropical diseases due to their low prevalence and still lower disease severity as compared to falciparum malaria [3,5,6]. The clinical features due to P. malariae and P. ovale spp. are thus poorly known, as very few reported studies are available compared to other human Plasmodium spp. [7]. P. ovale spp. are recognized to generally induce weak parasitaemia in the infected host, are very often responsible for relapses and sometimes have also been identified in certain renal complications [8]. In recent years, a steady rise in malaria infection cases associated with P. ovale spp. and P. malariae has been observed in many parts of sub-Saharan Africa and South America, with their prevalence being around 10% and 20%, respectively, in both symptomatic and asymptomatic malaria infections, causing concern for elimination programmes [9][10][11][12][13][14]. This increase in prevalence underscores the need for attention towards them, and therefore knowledge about their demographic distribution and associated morbidity is essential for understanding their epidemiology. In this review article, we have focused on estimating the disease burden and the epidemiological aspects of malaria infections caused by P. malariae and P. ovale spp., highlighting the need for increased attention towards these species for successful malaria elimination programmes.

Search strategy
The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines were used to select studies for this review [15], and six databases (PubMed, WHOLIS, Wiley Library, ScienceDirect, Web of Science and Google Scholar) were explored to access articles (Additional file 1). Although the search strategy differs slightly from one database to another, a systematic computerized strategy was adopted using MeSH terms and in combination with Boolean operators: ["malaria" OR "Plasmodium" AND ("malariae" OR "ovale" OR "curtisi" OR "wallikeri") AND ("prevalence" OR "epidemiology" OR "proportion" OR "frequency") AND ("PCR" OR "molecular")] (Additional file 2). In addition, a manual Google search and filtering of the reference lists of included studies were performed to access additional articles. The search included articles in both English and French languages published between January 2000 and December 2020.

Selection of studies and data extraction
All epidemiological studies that reported the prevalence of P. malariae and P. ovale spp. using molecular techniques as a species identification method were considered in this review. Only research articles published in peer-reviewed journals were included, and studies with no clear epidemiological data on Plasmodium spp. were excluded from the review. All articles were imported into Zotero bibliographic reference management software, and duplicate files were deleted. A data extraction form with the name of the first author, year of publication, period of data collection, geographical location (continent and country), target group (subject and subject status), type of study, sample size, sampling technique, diagnostic method, target gene and a positive result for each human Plasmodium species was prepared in a Microsoft Excel spreadsheet. The data files extracted independently by the first two authors were systematically evaluated for any inconsistencies, and any further discrepancies were resolved after discussion with the team supervisor.

Evaluation of criteria for inclusion of studies
The quality and risk of bias in included studies were assessed using an adapted version of the Joanna Briggs Institute (JBI) prevalence study assessment tool [16]. The evaluation criteria were based on eight parameters: (i) appropriate sample frame, (ii) appropriate sampling technique, (iii) adequate sample size, (iv) description of the study subject and settings in detail, (v) sufficient data analysis, (vi) use of validated methods for the conditions identified, (vii) valid measurement of the condition for all participants and (viii) use of appropriate statistical analysis. For each evaluation item, a score of 1 was awarded if found in the article; zero was given if not, and no score if the evaluation tool was unclear or not applicable to the study. Studies that received a score ≥ 5 were considered acceptable for final inclusion in the study (Additional file 3: Table S1).

Data analysis
Data analysis was performed with Metaprop command [17] in Stata 16.1 software (StataCorp LLC, USA). The combined prevalence and 95% confidence intervals (CIs) were computed using a random-effects model with Freeman-Tukey double arcsine transformation and restricted maximum likelihood method for stabilization and estimating the covariance. Heterogeneity between studies was assessed using I 2 , whose values of 25%, 50% and 75% were considered as low, moderate and high degrees of heterogeneity, respectively. The risk difference (RD) was performed to compare the occurrences of P. malariae/P. ovale spp. in mono-infections and coinfections with other species using OpenMeta[Analyst] software. Publication bias was also assessed using a funnel diagram by visually inspecting the plots and exact Egger's test for small-study effects. Possible sources of heterogeneity were assessed using subgroup analysis and meta-regression. We computed a subgroup analysis including only studies conducted in defined groups: (i) pregnant women, (ii) children ≤ 15 years old and (iii) participants > 15 years old. In the pregnant women group, when several samples were taken (peripheral blood, umbilical cord and/or placenta), only the results of the peripheral blood were used in the meta-analysis. A meta-regression with the sample size, date of publication and period of data collection as a covariate was performed to evaluate whether they were confounders of the P. malaria and/or P. ovale spp. pooled prevalence. All statistical analyses were considered statistically significant for values of p < 0.05.

Study selection and identification
A total of 3297 citations were screened initially, where 3278 were obtained from the six selected databases and 19 others through the manual Google search for inclusion in this review study. Five hundred and fortysix were duplicates and hence deleted. From the 2751 remaining articles, 2155 were excluded after reviewing the abstract. and a further 451 studies were excluded for one of the following reasons: (i) no access to the full text (n = 32), (ii) article in a language other than English or French (n = 4), (iii) epidemiological data on imported malaria cases (n = 46), (iv) comparative studies or diagnosis made by a new method (n = 19) and (v) not related to our study framework (n = 350). Of the remaining 145 articles, 32 were excluded because of their methodology or a lack of epidemiological information (Additional file 3: Table S2). Finally, 113 studies involving 195,065 subjects were considered for the systematic review (Fig. 1). Ninety-five and 102 studies were included in quantitative analysis of P. ovale spp. and P. malariae prevalence, respectively. The characteristics of the studies included in this review are summarized in Additional file 3: Table S3.

Characteristics of the included studies
Of the 113 studies included in our review, more than half (51.33%; n = 58) were published within the last 5 years, but no study was published in the year 2003 (Fig. 2a). The epidemiological studies included in this review covered five out of six pre-classified World Health Organization (WHO) health regions, with the African Region (AFR) accounting for half of the studies (57/113), followed by the Western Pacific (WPR) and South-East Asia regions (SEAR) with 21.24% (24/113) and 15.93% (18/113), respectively (Fig. 2b). The smallest and largest study samples had 47 and 17,765 participants, respectively, with an average sample size of 1726 (interquartile range [312-1724]). Studies from AFR (n = 17,765), the Eastern Mediterranean (EMR) (n = 16,075) and WPR (n = 13,980) reported the largest sample size (Additional file 3: Table S3). Forty-nine studies (43.36%) comprised both symptomatic and asymptomatic participants, whereas 26.55% and 30.09% included respectively asymptomatic and symptomatic participants only. Children under 15 years old were targeted in 10.62% of studies, mostly in AFR (n = 11); pregnant women were investigated in 7.08% in the African Region (n = 7) [18][19][20][21][22][23][24] and Region of the Americas (AMR) (n = 1) [25], and one study (0.88%) was conducted among 4570 women of childbearing age in the Democratic Republic of the Congo (DRC) [26] (Additional file 3: Table S3).

Prevalence among children, adults and pregnant women
On analysis, the P. malariae species was found to be equally distributed among adults (

Meta-regression and publication bias Evolution of prevalence over the years
The evolution of P. malariae and P. ovale spp. prevalence over the years was explored by meta-regression. Globally, a negative correlation between pooled prevalence and years of data collection was observed, though it was not significant for either P. malariae (Coef = −0.0131; p = 0.059) or P. ovale spp. (Coef = −0.0025; p = 0.431) (Fig. 5a, b). This observation was also the same for each region separately except for the WPR, where the P. malariae prevalence had declined significantly (Coef = −0.0212, p < 0.001) (Additional file 6).

Exploration of other potential confounders
According to the results, none of the variables of sample size, year of publication or period of data collection was a significant moderator in estimation of either P. malariae or P. ovale spp. prevalence; therefore, no evidence of publication bias was observed in the estimated prevalence due to these factors ( Table 2). In addition, the funnel plot does not indicate any evidence of publication bias for either of the two species. However, the exact Egger's test for small-study effect was significant for P. ovale spp.
(p = 0.046), where the small study size may be the reason for bias towards P. ovale spp. prevalence (Additional file 7).

Discussion
In the past two decades, WHO has reported a slight decline in malaria cases, from 238 million cases in 2000 to 229 million in 2019, with incidence declining from 363 to 225 cases per 1000 population through increased coverage of preventive and control measures (particularly long-lasting insecticide-treated nets, early diagnosis using rapid diagnostic tests, global improvement of WHO's policy development and dissemination processes for malaria control) [1, 39,40]. These have reflected in the regression of malaria and even its eradication in some countries during this period, notably in the United Arab Emirates, Morocco, Armenia, Argentina, Paraguay, Algeria, Uzbekistan, Sri Lanka, Turkmenistan and Kyrgyzstan [1]. This systematic review and meta-analysis estimated  Fig. 3 Forest plots of prevalence between asymptomatic and symptomatic participants: a P. malariae; b P. ovale spp. The prevalence of each study is represented by a spot and the horizontal bar centred by the prevalence gives the dispersion of the data the disease trend of P. malariae and P. ovale spp. from January 2000 to December 2020. The overall prevalence of P. malariae infections (2.01%) was higher as compared to P. ovale spp. (0.77%), with AFR being the most frequently affected for both species. Epidemiological studies on malaria conducted in the AMR, SEAR and EMR did not very often pay attention to P. malariae and/or P. ovale spp. due to their low  Plasmodium malariae estimated prevalence A Fig. 4 Forest plots of Plasmodium prevalence in studies conducted in specific groups (pregnant women, children under 15 and participants over 15 years old). a P. malariae prevalence, b P. ovale spp. prevalence. The prevalence of each study is represented by a spot and the horizontal bar centred by the prevalence gives the dispersion of the data prevalence/severity [12,[41][42][43][44][45][46], yet some studies report P. malariae prevalence of more than 40% in these regions [12,47]. This may be the reason for the underestimation of P. malariae and P. ovale spp. prevalence rates in these regions, as observed in this review. Negative correlations obtained in the meta-regression of prevalence as a function of data collection period indicate that the prevalence of both P. malariae and P. ovale spp. species have declined over the last 20 years, but not significantly. WHO has several strategic programmes to eliminate malaria, with a particular focus on P. falciparum and P. vivax malaria but none specific to P. malariae or P. ovale spp., which may have resulted in a low regression observed for these two species with a significant decline of P. falciparum and P. vivax prevalence recorded globally [39,48]. Gneme et al. [49], in a 4-year longitudinal study conducted in Burkina Faso from 2007 to 2010, revealed that the prevalence of P. malariae increased 15-fold and that of P. ovale spp. increased fourfold. Moreover, in the last 5 years, fairly high prevalence has been reported for P. ovale spp. [10,13] and P. malariae [9,10,12] (> 10% and 20%, respectively) in AFR and AMR due to the increased usage of diagnostic PCR assay [50]. The P. ovale spp. was reported in 12 out of 26 provinces of the DRC during 2007. Six years later, in 2013, the demographic and health survey revealed P. ovale spp. in 24/26 provinces of DRC, with an increased prevalence in several provinces [51]. This reflects a spread of P. ovale spp. that urgently needs to be addressed, as DRC is the second most affected country by malaria in the world [1]. Several studies have also reported an equal distribution of P. ovale subspecies in AFR [21,32,35,51]. High prevalence of asymptomatic malaria including P. malariae [9,34,42,[52][53][54] and P. ovale spp. [9,35,54,55] infections were reported in many studies. Our analysis indicates that both species showed the same distribution between symptomatic and asymptomatic subjects [56]. P. malaria and P. ovale spp. are often misidentified as P. falciparum and P. vivax, respectively, which complicates the routine diagnosis [8,57] and further leads to asymptomatic disease reservoirs [36] which could jeopardize malaria elimination programmes and therefore merits more attention.  Children and pregnant women are well known as vulnerable groups to malaria infection. Woldearegai et al. [10] reported in a Gabonese population that immunity to P. malariae builds up with age similarly to P. falciparum, but incompletely due to lower parasitaemia, which induces limited immune stimulation. In this study, subgroup analysis showed that P. malariae appears to have the same distribution in children, adults and pregnant women, whereas the highest P. ovale spp. prevalence among the subgroups was observed in pregnant women. Knowing that P. malariae has the same therapeutic protocol as P. falciparum [3,58], the use of intermittent preventive treatment in pregnancy (ITPp) contributes to the reduction of P. malariae burden in this group, unlike P. ovale spp., which can remain dormant in the liver (hypnozoites stage) for long periods. A cross-sectional study including 4570 women of childbearing age conducted by Taylor et al. [26] found that P. malariae was more prevalent in non-pregnant than pregnant women in the Wocba region (DRC). Unfortunately, this observation could not be linked with any factor, as the authors had no data on ITPp or antenatal care. On the other hand, Mitchell and colleagues reported P. ovale spp. prevalence twice as high in men (mainly miners) as in women, suggesting that male susceptibility was possibly due to the high exposure in their work environment. Therefore, the conclusions as to the risk groups for these two species remain controversial.
In the reported meta-analysis, the difference in risk of co-infections with another Plasmodium species reveals that both P. malariae and P. ovale spp. occurred more in co-infections, as reported earlier [9,10,13,[59][60][61][62]. The recurrence of the coexistence of two or more Plasmodium species can be explained by the densitydependent regulation model proposed by Bruce and Day [63,64]. This model suggests that when an individual carrying a Plasmodium species is reinfected by a second species, the parasitaemia due to the first species is down-regulated. However, when the species with the highest parasitaemia exceeds a certain threshold, the host immune response is triggered to control the infection. Once the infection is under control, the system is deactivated, promoting the growth of the second infecting species. This cycle is believed to be responsible for the persistence of several species in the host. In addition, it is also known that P. malariae and P. ovale spp. preferentially develop in mature and younger red blood cells respectively [65,66]. This differential target of invasion and no antagonism mechanism reported between these two species could explain their co-existence in an individual. Obviously, this requires further research, as not much data is available in this regard. Although not considered a significant threat to malaria elimination, the insufficient knowledge of the biological dynamics of P. malariae and P. ovale spp. should not be neglected because of their low prevalence and disease severity.
The strength of this study lies in the estimation of the prevalence rates of these two non-falciparum species based essentially on molecular techniques known to be the most sensitive diagnostic tool, and the global coverage of the WHO health regions in the last two decades, thus offering a global overview of P. malariae and P. ovale spp. malaria burden. However, the lack of whole information in some studies made it impossible to thoroughly analyse the data for identifying the environmental and biological factors that regulate the dynamics of these two species and which will be the subject of our future investigations.

Limitations
However, this study had certain limitations. First, the pooled prevalence data was estimated from studies conducted in different geographic areas and at different times, as reflected in the high heterogeneity observed. Second, the AFR had ten times as many papers as some of the regions included, which could be a source of imbalance in pooled prevalence data. Third, not all studies mentioned the characteristics of the study areas to permit us to explore other sources of heterogeneity. Fourth, only papers published in English or French were included in this review. Finally, field or laboratory working conditions that were not always clearly detailed and/ or sometimes differed could have generated several other confounding variables that affected our results.

Conclusion
This study summarizes the distribution and the global trend of P. malariae and P. ovale spp. prevalence in the last two decades. Based on our results, the prevalence of both species has declined slightly over this period, although high prevalence has been recorded in AFR and the AMR recently. Both species are equally prevalent in asymptomatic and symptomatic individuals of all age groups, which supports the interest of their inclusion in the new strategies for prevention, treatment and malaria elimination programmes.