Plasmodium vivax epidemiology in Ethiopia 2000-2020: A systematic review and meta-analysis

Background Ethiopia is one of the scarce African countries where Plasmodium vivax and P. falciparum co-exist. There has been no attempt to derive a robust prevalence estimate of P. vivax in the country although a clear understanding of the epidemiology of this parasite is essential for informed decisions. This systematic review and meta-analysis, therefore, is aimed to synthesize the available evidences on the distribution of P. vivax infection by different locations/regions, study years, eco-epidemiological zones, and study settings in Ethiopia. Methods This study was conducted in accordance with Preferred Reposting Items for Systematic Reviews and Meta Analyses (PRISMA) guidelines. Studies conducted and published over the last two decades (2000 to 2020) that reported an estimate of P. vivax prevalence in Ethiopia were included. The Cochrane Q (χ2) and the I2 tests were used to assess heterogeneity, and the funnel plot and Egger’s test were used to examine publication bias. A p-value of the χ2 test <0.05 and an I2 value >75% were considered presence of considerable heterogeneity. Random effect models were used to obtain pooled estimate of P. vivax infection prevalence. This study is registered with PROSPERO (International Prospective Register of Systematic Reviews): ID CRD42020201761. Results We screened 4,932 records and included 79 studies that enrolled 1,676,659 confirmed malaria cases, from which 548,214 (32.69%) were P. vivax infections and 1,116,581 (66.59%) were due to P. falciparum. The rest were due to mixed infections. The pooled estimate of P. vivax prevalence rate was 8.93% (95% CI: 7.98–9.88%) with significant heterogeneity (I2 = 100%, p<0.0001). Regional differences showed significant effects (p<0.0001, and I2 = 99.4%) on the pooled prevalence of P. vivax, while study years (before and after the scaling up of interventional activities) did not show significant differences (p = 0.9, I2 = 0%). Eco-epidemiological zones considered in the analysis did show a significant statistical effect (p<0.001, I2 = 78.5%) on the overall pooled estimate prevalence. Also, the study setting showed significant differences (p = 0.001, and I2 = 90.3%) on the overall prevalence, where significant reduction of P. vivax prevalence (4.67%, 95%CI: 1.41–7.93%, p<0.0001) was observed in studies conducted at the community level. The studies included in the review demonstrated lack of publication bias qualitatively (symmetrical funnel plot) and quantitatively [Egger’s test (coefficient) = -2.97, 95% CI: -15.06–9.13, p = 0.62]. Conclusion The estimated prevalence of P. vivax malaria in Ethiopia was 8.93% with P. vivax prevailing in the central west region of Ethiopia, but steadily extending to the western part of the country. Its distribution across the nation varies according to geographical location, study setting and study years.

Introduction Plasmodium vivax is one of the five human malaria parasites, with wider distribution across the globe [1]. It causes recurring malaria and affects a large number of populations globally [2]. Although it is widely accepted that the human P. vivax parasite has African origins [3], its presence in this continent has been unevenly distributed, and its clinical impacts are considered minor except in Eastern Africa [4]. Indeed, the horn of Africa (Ethiopia, Djibouti, Eritrea, and Somalia), South Sudan and the island of Madagascar seem to be the only countries where P. vivax is considered endemic and causes significant clinical disease in a stable manner, although reports from many other African countries confirm that the parasite does circulate beyond this region. Such a disparate distribution of clinical disease is probably linked to the higher prevalence in these countries (and its generalized absence in the rest of the continent) of Duffy positive individuals, given that this species is thought to require the Duffy receptor to invade reticulocytes and cause disease [5]. However, for the past decade, the increasing demonstration of P. vivax associated infections and diseases in Duffy-negative individuals from a variety of West African countries [6,7] confirm the underlying widespread presence of this species across other malaria-endemic regions of Africa, and the possibility that P. vivax has evolved to find an alternate ways of infecting the reticulocytes and causing disease [8]. Although this phenomenon is yet not widespread, it could further complicate achieving the current malaria elimination goals in the continent [7].
There are additional important knowledge gaps regarding P. vivax. The parasite's biology and its pathophysiology are still poorly understood, compared to that of P. falciparum. Current understanding of the hypnozoite and its basic biology remains elusive, and this is a critical gap that hampers current therapeutic and diagnostic strategies. Moreover, the early release of gametocytes to the bloodstream from the liver, even prior to the appearance of clinical symptoms, facilitates transmission, and obstructs control of this species. Such challenges significantly hamper current global P. vivax malarial control efforts, and calls for well-coordinated wider ranging research, surveillance and re-mapping of its global epidemiology [9].
Ethiopia accounts for 6% of the malaria cases globally, and about 12% of the global cases and deaths due to P. vivax [10]. The country has made significant efforts to control malaria since the introduction of dichlorodiphenyl-trichloroethane (DDT) as insecticide upon which the country based its indoor residual spraying (IRS) strategy back in 1959 [11,12]. Several attempts have been made to scale up major malaria interventional activities such as the distribution of insecticide treated bed nets (ITN), indoor residual spraying (IRS), and introduction of artemisinin-based combination therapy (ACT) starting from 2005 [13]. As a result of these concerted efforts, in areas with Annual Parasite Incidence (API) of > 100 per 1,000 population (high transmission), significant reductions of API (from 14.3 per 1,000 in 2013 to 6.4 in 2016 per 1,000 population) were documented [14]. However, in low transmission areas, the API appeared to increase from 22.5 to 37.4 per 1000 population from 2013 to 2016 [14].
In Ethiopia, where the burden of P. vivax seems to be slowly rivalling that of P. falciparum, no attempt has been made to derive a robust epidemiological review of the P. vivax data available in the country. Clear understanding of the distribution of P. vivax is essential for informed decisions on appropriate control strategies to be designed and implemented against this neglected species. Thus, the main aim of this review was to synthesize evidence on distribution of P. vivax infection among symptomatic and asymptomatic cases in Ethiopia.

Research design
The study was conducted according to Preferred Reposting Items for Systematic Reviews and Meta Analyses (PRISMA) guidelines. The protocol was registered at PROSPERO International prospective register of systematic reviews, with ID: CRD42020201761 (available at: https:// www.crd.york.ac.uk/PROSPERO/display_record.php?RecordID=201761).
(see S1 Table for the full detailed search strategies). Each search strategy was applied to articles published between 2000 and 2020. The last search was performed on 31 st December 2020. In addition, an effort was made to retrieve more information manually from African Journal Online (AJOL) indexed journals (n = 7). Grey literature and non-published data were not included in the review. Results from different database searches were exported to EndNote and then combined followed by trimming out of any duplicated data.

Eligibility criteria
Studies were eligible for inclusion if they were original publications describing the epidemiology of P. vivax in humans in Ethiopia. We included observational studies (cross-sectional and retrospective) written in any language and published over the last twenty years (from 1 st January 2000 to December 31st 2020). Studies conducted both in health facilities (i.e., health posts, health centers, and hospitals) and at the community level (i.e., villages, and schools) were included. Other data sources such as reviews, conference abstracts, commentaries, editorials, registered protocols for clinical trials, letters to the editor, personal opinions, non-human or in

PLOS NEGLECTED TROPICAL DISEASES
vitro studies, studies on other Plasmodium species and those with incomplete information (studies lacking data on prevalence of P. vivax) were excluded.

Study selection
Two authors (TK and KB) independently screened titles and abstracts of all records identified by the search strategy for potential inclusion in the review. Afterwards, full-text copies of articles deemed potentially relevant were retrieved and their eligibility was assessed. Disagreements between individual judgments were resolved through discussion. We listed all studies excluded after full-text assessment and reasons for the exclusion (S2 Table).

Data extraction
Two authors (TK and KB) used a data extraction form to independently extract data on study characteristics, including: type of study (facility or community based), age group, and presence or absence of symptoms. Additional information collected included study year (before or after the scale up of national malaria interventional activities) [14], geographical regions, diagnostic methods used, sample size, and the main characteristics of the population under study.
Outcome of interest was prevalence of P. vivax infection. P. vivax malaria diagnosis required parasitological confirmation irrespective of the methods used (optic microscopy, RDT, PCR, LAMP, ELISA, etc.). Original authors were contacted when further clarification and additional data were necessary.

Assessment of risk of bias in included studies
The risk of bias for each included study was assessed independently by two authors (TK and KB) using the Prevalence Critical Appraisal Instrument, designed to be used in systematic reviews addressing questions of prevalence, as described by Munn et al. [15]. This tool assesses the methodological quality of studies reporting prevalence data using ten critical appraisal criteria: sample representation of the target population, participant recruitment appropriateness, sample size adequacy, subjects and setting detailed description, enough coverage of the identified sample, objectivity and standardization in the measurement of the condition, reliability in the measurement of the condition, statistical analysis appropriateness, confounders/ subgroups/differences identification and accounting, and subpopulations identification using objective criteria. An overall low (�7/10), medium (between 5 and 7/10), high (<5/10) risk of bias level was assigned to each study.

Data synthesis and analysis
Data were analyzed using the Cochrane Review Manager (version 5.4) for qualitative and quantitative synthesis. Prevalence for each study was reported. For cases where prevalence was not reported, authors calculated it by dividing the event (P. vivax positive and/or in mixed infection) to the total population sampled in each study. Standard error of the mean (SE) for each study was calculated from the standard deviation obtained using the formula, StDev ¼ ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi pð1 À pÞ p where p is a proportion of the population with the event. Then, SE was calculated from the StDev using the formula, SE ¼ StDev ffi ffi ffi n p , where n is the sample size. Heterogeneity between studies was evaluated using Cochrane's Q (χ 2 ) and the I 2 tests. For the Cochrane's test, a p-value of the χ 2 test less than 0.05 was considered as significant statistical heterogeneity. I 2 values of 25%, 50% and 75% were assumed to represent low, medium, and high heterogeneity, respectively. Outliers that might cause heterogeneity and meta-coefficient were analyzed using Comprehensive Meta-analysis (CMA) software and presented using box plots (S1 Fig) and Table, respectively.
Subgroup analysis was conducted to investigate heterogeneity. Pre-specified subgroups potentially assumed to affect the overall prevalence estimate included: i) geographical location/regions (in Ethiopia there are currently ten regional states and two chartered cities), ii) study setting, iii) eco-epidemiological zones (altitude), and iv) study year. Likewise, due to high heterogeneity (I 2 > 75%, P < 0.05), random effects models were used for the pooled statistics. Forest plots were used to display point estimates and confidence intervals. Publication bias for studies included in the meta-analysis was assessed quantitatively using the Egger's test and qualitatively constructing funnels plot and looking for asymmetry. ArcGIS software version 10.0 was used to sketch a map for the distribution of P. vivax malaria in the country.

Study selection
A total of 4932 citations were initially identified. After the duplicates were excluded, 1841 unique citations were screened and assessed for eligibility. From the remaining 1841 screened at title/abstract level, 1715 records considered irrelevant for the purposes of the study were excluded. At the second phase of records assessment, a total of 126 eligible studies with available full text were thoroughly reviewed and a total of 72 articles (seven of them were comprised of a pair of an independent studies, which makes the total of studies 79) included for qualitative and quantitative meta-analysis, respectively (Fig 1). Detailed reasons for the 54 excluded studies are presented in S1 Table. Quality assessment of individual studies Across the 10 quality domains evaluated, the majority of the studies met five or more of the quality criteria. Most of the studies (n = 31) met 8 or more of the quality criteria assessed, and others (n = 26) met 5 to 7 of the quality criteria assessed for prevalence studies. Only 15 studies were rated below 5 for the quality assessment. The most common quality criteria not fulfilled by the studies were: poor statistical analysis such as failure to use reliable, valid and appropriate data analysis tools (n = 27), failure to identify confounders/differences accounting (n = 24) and unclear sample recruitment (n = 19). Most of the studies fulfilled the following quality criteria: contained adequate sample size (n = 64), described the study subjects and setting in detail (n = 62), and the data analyses were conducted with sufficient coverage of the identified samples (n = 69). Nine studies met all 10 quality assessment criteria. Twenty-eight studies were based on data extracted from patients' medical records accessed from health facilities. For such studies, some of the quality criteria such as defining target population, use of appropriate sampling techniques and standard data collection tools/methods were difficult to evaluate and were considered as not applicable (NA) (S3 Table).
Twenty-eight studies reported pooled prevalence data based on retrospective evaluations of 5-20 years' patient data collected from health facilities. The remaining 51 were cross sectional studies undertaken at health facilities (n = 60) or at the community level (n = 19). Malaria diagnosis relied on optic microscopy in the majority of studies (n = 60/79, 75.95%); with the remaining 19 studies using either only RDT (n = 3), microscopy plus RDT (n = 11), microscopy plus PCR (n = 2), a mix of the three techniques (microscopy, RDT and PCR; n = 3). Participants of most of the included studies (n = 59/79, 74.7%) were all-age groups populations, while 11 were from children and teenagers up to 15   . Ethiopia is a federal state https://en.wikipedia.org/wiki/ Federation subdivided into ethno-linguistically based regional states. There are currently ten regional states and two chartered cities. In line with this division, the studies reported data from the regions of Afar (n = 1), Amhara (n = 26), Benishangul (n = 3), Oromia (n = 18), Southern Nations, Nationalities and Peoples' Region (SNNPR) (n = 25), Tigray (n = 1), Harari (n = 1) and nationwide surveys of Ethiopia (n = 4). Accordingly, the majority of the malaria research reports (69/79, 87.34%) presented data from Amhara, Oromia and SNNPR. Based on the eco-epidemiological zones of malaria distribution, 22 studies were reported from areas with altitude <1500m (low lands with seasonal/intense transmission), 10 were from altitudes between 1500-1750m (high land fringe, high unstable transmission), 14 were from altitudes ranging between 1750-2000m (high land fringe, low unstable transmission), 7 studies were from districts with altitudes of 2000-2500m (highland, occasional epidemic) and 23 were from areas with mixed ecological zones (Table 1), and three studies without this information were excluded [48,49].

Main outcome of the meta-analysis
The overall random effects pooled prevalence rate of P. vivax (mono-infection and mixed infection with P. falciparum) in Ethiopia was 8.93% (95% CI: 7.98-9.88%), with a very high level of heterogeneity (I 2 = 100%, p<0.0001). Indeed, the prevalence of P. vivax across individual studies varied considerably [ranging from 0.25, n = 1/400 among all age groups in SNNPR [85] to 47.35%, n = 197/416 in all age groups in many sites throughout Ethiopia using 18r based nested PCR [74] (Fig 3).
The pooled prevalence of P. vivax in mono-infection was 7.98% (95% CI: 7.09-8.87%) with a very high level of heterogeneity (Fig 4) and prevalence of P. vivax in a mixed infection (P. vivax with P. falciparum) was 0.73% (95% CI: 0.65-0.82%). The prevalence reported in each study for mixed infection was also varied and ranged from 0.005% [51]  Regional variation showed significant effect on the estimated prevalence of P. vivax although there was high significant heterogeneity (I 2 = 100%, p<0.0001) within each of the three main regions (Amhara, Oromia and SNNPR). SNNPR is a region where significantly highest (10%, 95%CI: 8.46-11.54%) pooled prevalence of P. vivax is documented (S3 Fig).
Three studies (one of them contained a pair of studies) included in the review, which reported national/regional or more than one region prevalence were excluded from the locations/ region's analysis [58,86,87] The different eco-epidemiological zones considered in the meta-analysis did appear to significantly affect the pooled estimate prevalence of P. vivax (χ 2 = 18.65, df = 4, p = 0.0.01, I 2 = 78.5%). Moreover, some studies reported from the highlands with occasional malaria epidemic zones (2000-2500m) contributed to the observed high prevalence of P. vivax (9.80%, 95%CI: 6.73-12.87%) compared to other eco-epidemiological zones (S4 Fig).
There were significant study setting differences (facility and community) among the studies (χ 2 = 10.27, df = 1, p = 0.001, and I 2 = 90.3%). Being diagnosed and treated at the health facility (health centers, health posts and hospitals) significantly (10.44%, 95%CI: 9.09-11.79%, p<0.0001) affected the overall pooled prevalence of P. vivax, although there was substantial unexplained high heterogeneity within the studies conducted at both settings (I 2 = 100% for both). Hence, the validity of study setting effect estimate for each subgroup is uncertain as

PLOS NEGLECTED TROPICAL DISEASES
individual studies were inconsistent. However, a significant reduction in the prevalence of P. vivax (4.67%, 95%CI: 1.41-7.93%, p<0.0001) was observed in studies conducted at the population /community level (schools, and villages) (S5 Fig). Analysis of effects of study years on the pooled estimated prevalence of P. vivax revealed lack of statistically significant differences (p = 0.93, I 2 = 0%) within the subgroups (S6 Fig).

Meta-regression analysis
A meta-regression analysis was used to determine if sub-groups (geographical situation, altitudes of the study sites, years of the study and study settings) had an effect on the pooled prevalence of P. vivax in the country. Findings from this meta-regression analysis further confirmed the effect of the subgroups on the overall pooled P. vivax prevalence. Geographical situation of the studies (SNNPR region), study settings (study from health facilities compared to those from community), and studies reported from areas whose altitude ranges from 1500-1750m seemed to be associated with a significant increasing in the prevalence of P. vivax malaria in Ethiopia, but the remaining variables such as study year did not show significant effect on the pooled prevalence of P. vivax. Studies from altitude ranges from 2000 to 2500m showed comparatively higher prevalence of P, vivax next to altitude range from 1500-1750, although significant difference was not observed ( Table 2).

Discussion
This study aimed to review the overall prevalence of P. vivax malaria infections in Ethiopia.
For this purpose, any study that investigated the prevalence and epidemiology of malaria in the country, and which contained detailed data on P. vivax was included. The overall pooled prevalence of P. vivax malaria (mono-infection or mixed infection among symptomatic and asymptomatic patients) in Ethiopia was 8.93% (95% CI: 7.98-9.88%). Prevalence among P. vivax mono-infection alone was 7.98% (95% CI: 7.09-8.87%). These figures are much higher than the predicted endemicity values of P. vivax prevalence for Madagascar and Ethiopia, and parts of South Sudan and Somalia, which rarely exceed 2% [87]. Typically, the P. vivax parasite load in peripheral blood is very low as compared to P. falciparum, often hindering its diagnosis using conventional optic microscopy [88]. However, such low-level parasitemias are sufficient to act as reservoirs and sustain transmission of the parasite [89]. Although microscopy is still the gold standard tool for malaria diagnosis in Ethiopia, a more accurate approach for diagnosis would require the use of more sensitive techniques such as PCR or LAMP, capable of detecting submicroscopic carriage and mixed infections in areas where the two main parasites (P. falciparum and P. vivax) co-exist [90]. Given that most of the studies included in this review used microscopy as the chosen diagnostic tool, it is likely that the reported prevalence rates are an underestimate of the true prevalence of this parasite. Ethiopia has variable topographic features that govern the distribution of malaria infection. Generally, it is agreed that malaria is endemic in areas with altitude lower than1500m (lowlands with seasonal/intense transmission) and rare in areas above 2000m (highland with occasional epidemic) [91]. However, in contrast to the general assumption, some studies reporting data from the highlands known for occasional malaria epidemics were found to contribute for a higher prevalence (9.80%, 95%CI: 6.73-12.87%) of P. vivax. This might be attributed to its survival ability in colder climate than other Plasmodium species [92]. A recent nationwide malaria epidemiological and interventional survey report confirms this finding, establishing the expansion of malaria to areas with altitude higher than 2000m [14], which were previously considered malaria free zones [93] and re-classified them as with moderate annual parasite incidence (APIs). The same report further indicated this as a new risk factor interfering with the current national malaria interventional activities [14]. A sero-prevalence study further strengthened the lack of significant differences in the transmission of P. vivax due to altitudinal variation (below or above 2000m) [93]. Rather, P. vivax showed direct relation with increasing elevation among children aged <5 years and high sero-positivity (20.9, 95% CI: 17.4-24.9) was observed at higher elevations [93]. The increasing evidence on the transmission of P. vivax in the areas traditionally considered as malaria free is an indication of the expansion of malaria transmission in Ethiopia to higher altitude settings. This expansion might be attributed to different developmental plans such as dam constructions, and the use of river water for irrigation purposes, deforestation, population pressures, and lack of appropriate environmental management system [86,94], which could cause local environmental modifications contributing to the creation of new suitable vector breeding sites or expansion of mosquito's habitat to nonendemic regions; besides changing human settlement pattern [95]. Malaria is one of the most climate sensitive diseases [96,97] with significant associations between malaria incidence and temperature [96], relative humidity [97,98] and rainfall [99], all of which do play a significant role in malaria transmission, which makes the vector controlling efforts very challenging. In addition, there are several Anopheles species with some different complexes, thus facilitating transmission into different ecological niches [100]. Furthermore, unlike other plasmodium species, P. vivax is capable of undergoing sporogonic development in the mosquito at lower temperatures [101] and able to expand to the highland areas. Growing evidence on P. vivax malaria distribution across other areas of Sub-Saharan Africa has further revealed that P. vivax appears to become proportionally more significant where overall malaria prevalence is lower [9].
Regional variation on P. vivax malaria prevalence was observed in the current review. In very recent years, significant reduction in P. vivax malaria burden has been predominantly observed in the Oromia region, as compared to the other regions [19,72]. According to the National Strategic Plan for Malaria Prevention, Control and Elimination in Ethiopia, the malaria burden was significantly reduced over three survey years (2007, 2011 and 2015) with 0.3% nationwide prevalence in the year 2015 [90]. This figure is relatively lower than reports made from other regions including SNNPR (0.5%), Amhara (0.8%), Benshangul (2.7%) and Gambella (6%) in the same year [90].
Compared to the national report, the prevalence of P. vivax malaria infection reported in the current review is much higher. This is due to the fact that the national report was the overall national malaria prevalence, which included only recent data (after malarial morbidity and mortality burden started decreasing) from all malaria transmission settings (low, middle, and high). But, this review only focused on prevalence of P. vivax malaria infection and included almost all studies conducted at high malaria transmission areas, and the prevalence data of 20 years. The recent national sero-prevalence analysis by region supports this finding, with lower P. vivax sero-prevalence documented in Oromia than in Amhara (36.7% (95% CI: 30.0-44.1) and SNNPR regions [92], although the detected antibodies might not correspond adequately to the existing infection prevalence.
Following the rise in malaria prevalence as observed in the year 2010/11, the deployment of malaria interventions already initiated in Ethiopia was boosted. This included the distribution of free ITN, IRS, and RDTs as a supplement for malaria diagnosis in remote areas, and the scale-up of ACT deployment and training of health extension workers [102]. As a result, the overall national malaria burden decreased from 0.5% prevalence in 2011 to 0.3% in 2015 [90]. Our metaanalysis on studies whose survey years were before and after the scaling-up of national malaria intervention activities did not show significant effect on the pooled estimated prevalence of P. vivax in Ethiopia. However, results from meta-regression indicated that prevalence of P. vivax observed after the scaling up of the interventional activities in Ethiopia, showed significant reduction. This finding is in agreement with the global P. vivax malaria burden reduction observed (41.6% reduction from 2000 to 2017) in most endemic areas [103]. Although the trend showed a declining pattern, burden due to P. vivax in Ethiopia appears considerable, and will cause enormous challenges, calling for careful regular surveillance by concerned bodies. Mainly it's apparent complex parasite biology, pathophysiology, treatment response, the raising problem of Duffy negative individuals that are now infected by P. vivax and transmission patterns [104] will make its future eradication goal very challenging. In addition, the hypnozoite's dormant liver stages, responsible for the potential repeated relapses that can occur within weeks, months, or many years after the initial inoculation, blur our current understanding of P. vivax epidemiology, and will not be affected unless specific radical cure is conducted [102]. In the absence of such antihypnozoite drugs, the current first line drugs used in Ethiopia for P.vivax malaria, be it chloroquine or other artemisinin based-combination therapies, will not affect the liver stage hypnozoites [9], thus hindering its adequate control. In addition, ITN and IRSs currently in use might not be efficient in completely preventing new infection, in general, and the relapse from liver stages in particular [9], mosquito species that transmit P. vivax bite mostly outdoors and which also changed its biting time from midnight to dawn [105]. Some populations of An. arabiensis were reported to even avoid fatal insecticide exposure [106,107].

Strengths and limitations of the study
To the best of our knowledge, this is the first detailed systematic review and meta-analysis of only P. vivax epidemiology in Ethiopia that included facility and community level studies. A recent systematic review and meta-analysis by Deresse and Girma, [108] assessed (using 35 studies) the prevalence of P. falciparum and P. vivax in Ethiopia and found 25.8% prevalence all together. Its main objective was to show a general picture on malaria prevalence in Ethiopia. Hence P. vivax prevalence/epidemiology was not uniquely reviewed, analyzed or presented separately in the study. Furthermore, the study didn't include the major databases such as Web of Science, Scopus, and EMBASE, but only retrieved articles from PubMed and Google scholar. In addition, it did not assess the role of subgroups such as location, eco-epidemiological zones, study setting and survey years, on the overall pooled prevalence of malaria, in general, and P. vivax in particular. The omission of subgroups appears to have significant impact, given that these subgroups showed a significant role on the estimated prevalence of P. vivax in our analysis. Hence, the strength of this review is the fact that it included many other new studies to date (n = 44) on P. vivax in Ethiopia besides the 35 studies included in the previous review and portrayed the epidemiological distribution of P. vivax nationwide [108] The major limitation of this review was that about one third of the included studies depended on data extracted from retrospective medical case records, reviewed to investigate the prevalence and trends of malaria. Although case record reviews are the most universally used method for prevalence studies, it is often challenging to obtain, in a standardized way, all required data about the individual patient, including socio-demographic and clinical data, how target groups were identified, recruited and the exact diagnostic tools used at the time of enrollment of each participant. In addition, for some of the studies included in the review, their main objective was not set to assess the prevalence or geographical distribution or epidemiological trends of malaria. Some were designed to show association between malaria prevalence and ABO blood groups/helminthic infection/HIV infection/ ITN utilization /hematological profile of malaria patients/ drug efficacy evaluation against P. vivax/or comparative evaluation of different malaria diagnostic tests or tools (microscopy Vs PCR). Data from this kind of studies often don't allow an adequate evaluation of the quality criteria set for prevalence/observational studies. Thus, they were included in the review only if they contained data on prevalence of malaria and different Plasmodium species. Moreover, significant heterogeneity of the eligible studies observed in this review may require further analysis. Finally, the exclusion of unpublished studies as well as interventional studies may lead, potentially, to loss of substantial data.

Conclusions
The overall estimated prevalence of P. vivax was 8.93% (95%CI: 7.98-9.88). Most of the studies included in the current review met the quality criteria and there was no publication bias. This parasite has historically been widely distributed in the central west region of Ethiopia, and is now steadily extending to the North West and South West regions of the country. Oromia, Amhara and SNNPR are the three major regions where P. vivax has spread predominantly with wide-ranging prevalence. P. vivax epidemiology has shown the trend of expansion to the highland, causing occasional malaria epidemics, although the existing deployed interventions seem to have an impact on prevalence of this parasite.
Supporting information S1