Introduction

While Plasmodium falciparum is the most prevalent malaria parasite in the World Health Organization African Region and accounted for 99.7% of estimated malaria cases in 20181, there are increasing reports of P. vivax infection across Africa2,3. P. vivax infection of human erythrocytes requires the presence of a glycoprotein on the surface of red bloods, the Duffy blood group antigen or the Duffy Antigen Receptor for Chemokines (DARC)4,5. DARC is also the receptor for the simian malarial parasite, Plasmodium knowlesi6. DARC binds to P. vivax Duffy binding protein (PvDBP) before it invade erythrocytes7,8. The Duffy blood group is expressed by the FY gene on chromosome 1, and is genotyped as FY (a), FY (b), FY (a)ES, and FY (b)ES9. Duffy phenotypes, including Fy(a + b +), Fy(a + b −), and Fy(a − b +) are Duffy-positive phenotypes, while Fy(a − b −) or FY (a)ES(b)ES are Duffy-negative phenotypes. The Fy(a − b −) phenotype is caused by homozygosity of the FY allele carrying a point mutation at 67T > C (rs2814778) which prevents Duffy antigen expression in red blood cells10.

The Duffy-negative phenotype is highly predominant in sub-Saharan African populations, with high phenotype median frequencies of 98%–100% in west, mid, and south-eastern regions5. Recent studies reported that Duffy-negative individuals have a risk of P. vivax infection11,12. It was also postulated that P. vivax infections were passed back and forth between Duffy-positive and Duffy-negative individuals by P. vivax-infected mosquitoes parasitizing Duffy-positive individuals and transmitting parasites to Duffy-negative individuals13. As P. vivax infection can lead to severe malaria with poor outcomes14, a better understanding of P. vivax infection occurrence and risk among Duffy-negative individuals is required to guide further epidemiological research in Africa. Therefore, using a meta-analysis approach, we investigated P. vivax infection prevalence among Duffy-negative individuals and assessed the risk of infection among these individuals when compared with Duffy-positive individuals.

Methods

Protocol and registration

This study followed Preferred Reporting Items for Systematic review and Meta-Analyses guidelines15. The review was registered at The International Prospective Register of Systematic Reviews website (ID: CRD42021240202).

Search strategy

Literature searches were conducted using medical subject headings in the National Library of Medicine and terms related to P. vivax malaria and Duffy status. The following search terms were used: “DBP” OR “D binding protein” OR “D-element-binding protein” OR “DBP transcription factor” OR “D-site binding protein.” Search terms are shown (Table S1). Medline, Web of Science, and Scopus were searched from the February 22, 2021 to the January 31, 2022. Additional searches of reference lists and review articles were also performed to ensure literature saturation.

Eligibility criteria

Cross-sectional, cohort, and case–control studies were considered if they reported P. vivax infections among Duffy-negative individuals. P. vivax infection was confirmed by microscopic or molecular analysis. Duffy genotypes or phenotypes were characterized by polymerase chain reaction-restriction fragment length polymorphisms, with and without sequencing. Only articles in English were included. The following articles were excluded: no Duffy-negative individuals among P. vivax cases, genetic analysis of the Duffy protein, no report on Duffy status, case reports and case series, experimental studies, clinical trials, and studies from which data could not be extracted.

Study selection

Study selection was performed in Endnote (Version X8, Clarivate Analytics, USA) by two authors (PW and MK). Discrepancies between authors on study selection were resolved by consensus and discussion with a third author (KUK). After retrieving articles, duplicated articles were removed. The remaining articles were title and abstract screened, after which irrelevant studies were excluded. The remaining article texts were examined according to eligibility criteria. All excluded articles were assigned appropriate reasons. Selected articles were further extracted using a standardized pilot datasheet.

Data extraction

The standardized pilot datasheet included the following: first author name, year of publication, study site, year the study was conducted, participants, age, gender, number of patients with malaria, number of P. vivax cases, number of P. vivax infections among Duffy-negative individuals, number of P. vivax infections among Duffy-positive individuals, malaria identification methods, and Duffy status. Two authors (PW and MK) independently collected these data. Disagreements over data extraction were resolved by discussion. Data were randomly checked by a third author (FRM) for completeness, plausibility, and integrity, before data was processed.

Study quality

The methodological quality of selected studies was evaluated using the Joanna Briggs Institute (JBI) Critical Appraisal Tools for assessing cross-sectional, case–control, and cohort studies16. The tool for cross-sectional studies comprised eight checklists, whereas 10 and 11 were used for case–control and cohort studies, respectively. Studies with > 75%, 50%, and ≤ 50% scores indicated high, moderate, or low quality, respectively. Study quality was assessed by two authors (PW and MK).

Study outcomes

The primary study outcome was the pooled prevalence of P. vivax infection among Duffy-negative individuals. The secondary outcome was the odds ratio (OR) and 95% confidence interval (CI) of P. vivax infection among Duffy-negative individuals when compared with Duffy-positive individuals.

Data processing

Primary and secondary study outcomes were both estimated using the random-effects model. This model was used in the presence of heterogeneity of the effect estimates (ES) (pooled prevalence or OR); meanwhile, the fixed-effects model was used in the absence of heterogeneity of the ES. The results from individual studies were graphically represented on forest plots. Heterogeneity among studies was assessed using Cochrane Q and I2 statistics. A Cochrane Q p < 0.05 indicated significant heterogeneity among studies. I2 statistics were used to quantify heterogeneity; I2 > 50% indicated substantial heterogeneity. If heterogeneity existed, the random-effects model was used for pooled the pooled prevalence and OR, and if no heterogeneity was observed, the fixed-effects model was used for pooled the pooled prevalence and OR. Meta-regression analysis was performed to determine the source(s) of heterogeneity of ES (pooled prevalence, OR) among studies. If the source(s) of heterogeneity was identified, a subgroup analysis was conducted. We performed sensitivity analysis of the pooled prevalence and the odds of infection between Duffy-negative individuals using the fixed-effects model to determine the robustness of our meta-analysis results.

Publication bias

Publication bias was assessed by visualizing funnel plot asymmetry and the Egger’s test. Funnel plot asymmetry indicated publication bias. A significant Egger’s test (p < 0.05) indicated that funnel plot asymmetry was due to a small study effect. If the funnel plot was asymmetrical (by visualization or a significant Egger’s test), a contour-enhanced funnel plot was generated to identify if funnel plot asymmetry was due to publication bias or other causes.

Results

Search results

Of 1593 retrieved articles, 806 were retained after duplicated articles were removed. After screening title and abstracts of 787 articles, 707 were excluded due to irrelevance (Fig. 1). Thus, 80 articles were examined for full texts and 54 excluded due to the following reasons: nine full texts were unavailable, nine texts reported no P. vivax cases in Duffy-negative patients, four texts had no Duffy-negative patients with P. vivax infection, four texts indicated prior exposure to malaria and Duffy status, four texts reported Duffy gene polymorphisms and P. vivax infection, four texts had no Duffy data, three texts had Duffy and P. vivax data which could not be extracted, two reported DBP polymorphisms and P. vivax infection, two texts used the same participants, two texts were in vitro studies, two had no P. vivax cases, two reported a Duffy mutation and P. vivax infection, one text was a P. vivax genomic analysis, one text reported P. vivax (1 case) in Duffy-positive patients, one was a letter to the editor, one reported Duffy-negative heterozygotes and a P. vivax infection, one reported Duffy status in non-malaria patients, one was a mosquito-infectivity study, and one was an editorial. Thus, 26 studies17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42 met eligibility criteria, however, one study43 was identified from the bibliography of a study, therefore 27 studies17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43 met eligibility criteria and were included.

Figure 1
figure 1

Study flow diagram demonstrating study selection process.

Full size image

Study characteristics

Study characteristics are shown (Table 1). All were published between 2006–2021 and almost all (24/27, 88.9%) reported P. vivax infection among Duffy-negative individuals in Africa. Three studies20,21,31 were conducted in South America (2/27, 7.4%) and Asia (1/27, 3.7%). Of the 24 African studies, six were conducted in East Africa (Ethiopia28,43, Madagascar25,29, Kenya40, and Ethiopia41), seven in Mid Africa (Democratic Republic of Congo19, Cameroon22,23,32,33,39, Angola, and Equatorial Guinea30), seven in West Africa (Mauritania24,42, Nigeria36,37, Senegal34, Mali35, and Benin38), two in North Africa (Sudan17,18), one in North and East Africa (Ethiopia and Sudan26), and one in Ethiopia/Botswana/Sudan27. Twenty-two articles were cross-sectional studies (22/27, 81.5%), two were case-controls12,40, and one was a cohort study35. The geographical distribution of studies is shown (Fig. 2).

Table 1 Characteristics of the included studies.
Full size table
Figure 2
figure 2

Distribution of included studies on P. vivax infection among Duffy-negative individuals. Map was sourced and modified from https://mapchart.net/world.html by authors. Authors were allowed to use, edit and modify any map created with mapchart.net for publication freely by adding the reference to mapchart.net in publication.

Full size image

Study quality

Study quality was assessed using the JBI Critical Appraisal Tool (Table S2). Eighteen studies18,19,21,23,25,26,27,28,29,32,34,35,36,37,38,39,41,42 were high-quality, while nine17,20,22,24,30,31,33,40,43 were of moderate quality.

The prevalence of P. vivax infection among Duffy-negative individuals

Twenty-seven studies17,18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36,37,38,39,40,41,42,43 reported P. vivax infection among Duffy-negative individuals. Of these, 1117,22,30,32,33,34,35,36,37,38,40 reported that all P. vivax infection cases were Duffy-negative (100%). These studies were conducted in West Africa (Nigeria36,37, Senegal34, Mali35, Benin38), Mid Africa (Cameroon22,32,33, Angola and Equatorial Guinea30), North Africa (Sudan17), and East Africa (Kenya40).

Fourteen studies18,19,20,21,23,24,25,27,28,29,39,41,42,43, conducted in 16 areas and reporting P. vivax infection prevalence among Duffy-negative individuals, were included in the pooled prevalence meta-analysis. These results showed that the pooled prevalence was 25% (95% CI − 3%–53%, I2 = 99.96%, Fig. 3). Due to high heterogeneity in studies reporting this prevalence, a meta-regression analysis of the continent as a covariate was performed to test if it (the continent) was a source of heterogeneity. These results showed that the continent covariate was indeed a source of heterogeneity in the pooled prevalence (p = 0.013), therefore, further subgroup continent analyses were performed.

Figure 3
figure 3

Forrest plot demonstrated the pooled prevalence of P. vivax infection among Duffy negative individuals. ES prevalence estimate, CI confidence interval.

Full size image

These results indicated that the highest prevalence of P. vivax infection among Duffy-negative individuals was identified in a Southern African study (Botswana, 86%, 95% CI 65%–95%)27, followed by Mid Africa (61%, 95% CI 6%–115%, I2 = 99.59%, three studies 19,23,39), and North Africa (13%, 95% CI 9%–18%, I2 = 100%, two studies18,27). However, a low prevalence was reported in an East African study [6%, 95% CI 3%–9%, I2 = 83.96%, five studies (six study areas)25,27,29,41,43], followed by West Africa (4%, 95% CI 1%–8%, I2 = 96.79%, two studies24,42). But the lowest prevalence was reported in a South American study (Brazil, 1%, 95% CI 0%–2%, I2 = 99.8%, two studies20,21) (Fig. 4).

Figure 4
figure 4

Forrest plot demonstrated the pooled prevalence of P. vivax infection among Duffy negative individuals stratified by continents. ES prevalence estimate, CI confidence interval.

Full size image

The odds of P. vivax infection among Duffy-negative individuals

The odds of P. vivax infection among Duffy-negative individuals when compared with Duffy-positive individuals were estimated using data from 11 studies12,18,20,23,25,31,39,41,44,45,46. Results of individual study showed that Duffy-negativity was a protective factor for P. vivax infection in six studies17,25,41,44,45,46. These studies were conducted in Sudan18, Madagascar25,45, Ethiopia41,44, and Mauritania46. Only one study conducted outside Africa (Brazil) demonstrated a higher risk of P. vivax infection among Duffy-negative individuals20. No differences in infection risk were identified in four studies from Cameroon23,39, Ethiopia and Sudan12, and Iran31. Overall, our pooled analysis of 11 studies demonstrated a decreased odds of P. vivax infection among Duffy-negative individuals (p = 0.009, pooled OR 0.46, 95% CI 0.26–0.82, I2 = 80.8%, 11 studies, Fig. 5).

Figure 5
figure 5

Forrest plot demonstrated the odd of P. vivax infection among Duffy negative individuals. OR odds ratio, CI confidence interval.

Full size image

Due to a high degree of heterogeneity in some studies, a meta-regression analysis of country, continent, and study design as covariates, was performed to test if covariates were heterogeneity sources of the pooled OR; continent was identified as a heterogeneity source (p = 0.027), whereas, country and study design were not heterogeneity sources of the pooled OR (p = 0.06 and p = 0.188, respectively).

Subgroup continent analysis showed that the decreased odds of P. vivax infection among Duffy-negative individuals were identified in studies in North Africa (OR 0.50, 95% CI 0.32–0.80)18, East Africa (pooled OR 0.24, 95% CI 0.11–0.52, four studies25,28,29,41), and West Africa (OR 0.40, 95% CI 0–0.27)42. Also, the increased odds of P. vivax infection among Duffy-negative individuals were identified in a South American study (OR 6.36, 95% CI 1.23–32.88)20. Other studies from Mid Africa23,39, North and East Africa26, and Asia31 showed no differences in the odds of infection between Duffy-negative and Duffy-positive individuals (Fig. 6).

Figure 6
figure 6

Forrest plot demonstrated the odd of P. vivax infection among Duffy negative individuals stratified by continents. OR odds ratio, CI confidence interval l.

Full size image

Sensitivity analysis

The sensitivity analysis showed that the pooled prevalence was 45% (95% CI 44%–45%, 14 studies in 16 areas, Supplementary Fig. 1). The decreased odds of infection between Duffy-negative individuals when compared with Duffy-positive individuals was p = 0.009, OR 0.46, 95% CI 0.26–0.82, 11 studies (Supplementary Fig. 2).

Publication bias

A funnel plot between ES (OR) and standard error of the logES of 11 studies showed a symmetrical funnel plot (Fig. 7). Egger’s test results showed no small study effects (p = 0.188). Contour-enhanced funnel plot analyses were performed to identify if funnel plot asymmetry was due to publication bias or other causes. These results showed that the ES’s were distributed in both significant and non-significant areas, thereby suggesting funnel plot asymmetry was due to other causes (e.g., heterogeneity in the OR between studies) (Fig. 8).

Figure 7
figure 7

The funnel plot between odds ratio (OR) and standard error (se) of the logOR of the 11 studies demonstrated that the funnel plot was asymmetry. OR odds ratio, se standard error.

Full size image
Figure 8
figure 8

Contour-enhanced funnel plot demonstrated that the effect estimates were distributed in both significance and non-significance areas indicating that the funnel plot asymmetry was due to other causes.

Full size image

Discussion

Duffy-negative individuals are typically resistant to P. vivax infection; however, a recent study showed that the Duffy-negative antigen was no longer a barrier to such infections30. In our review, we collated 27 studies showing P. vivax infection among Duffy-negative individuals in Africa, including Cameroon, Ethiopia, Sudan, Botswana, Nigeria, Madagascar, Angola, Benin, Kenya, Mali, Mauritania, Democratic Republic of the Congo, and Senegal. Moreover, three studies20,21,31 reported infections among Duffy-negative individuals in South America (Brazil)20,21 and Asia (Iran)31.

Our qualitative analyses showed that several studies17,22,30,32,33,34,35,36,37,38,40 reported that 100% P. vivax infection occurred in Duffy-negative individuals. In addition, our quantitative analyses (meta-analyses) showed that the pooled prevalence of infection among Duffy-negative individuals was 25%, with a high heterogeneity across studies. These finding confirmed data from previous studies and supported the hypothesis that Duffy-negativity was no longer protective against P. vivax infection. Nevertheless, a high prevalence of infection among Duffy-negative individuals was observed in West Africa34,35,36,37,38), Mid Africa19,22,23,30,32,33,39), North Africa17,18,27, East Africa40, and Southern Africa27. Our meta-analysis results showed that Duffy-negativity was protective against P. vivax infection in individuals from East Africa25,28,29,41, although several reports have documented about the infection of P. vivax in Duffy- negative individuals. Our forest plot demonstrated the increased odds of P. vivax infection among Duffy-negative individuals in studies outside Africa, such as South America. This was likely caused by a low sample size, as the authors suggested P. vivax infections were not significantly different between Duffy-positive and Duffy-negative individuals20.

Several mechanisms have been postulated for P. vivax infections among Duffy-negative individuals. (1) Duffy-positive individuals may act as P. vivax reservoirs and facilitate parasite infection of Duffy-negative hepatocytes, thereby selecting new P. vivax strains which invade Duffy-negative erythrocytes via Duffy-independent mechanisms45. (2) P. vivax evolution for host selection may have occurred in Africa due to ideal temperatures and highly competent transmission vectors17. (3) In Africa, increased vector capacity to transmit other P. vivax malaria parasites such as Anopheles gambiae and An. Arabiensis has been observed40,47. Demographic factors and a high population density of young age groups may have contributed to a higher entomological inoculation rate, and contributed to P. vivax infection in Duffy-negative individuals, similar to P. falciparum infection12,48. (4) Parasite adaptation may have occurred for P. knowlesi infection rates, potentially facilitating the zoonotic transmission of specific P. vivax strains in Duffy-negative individuals, resulting from long exposure to P. vivax infections in African populations. In studies on simian malaria parasites requiring the Duffy protein antigen for erythrocyte invasion, P. knowlesi invaded Duffy-negative erythrocytes, suggesting a Duffy-independent P. knowlesi infection mechanism49. (5) P. vivax can hide in the bone marrow of Duffy-negative hosts and persist as low parasitemic, asymptomatic infections50. (6) Difference in latitude in some areas could affect P. vivax transmission, e.g., higher altitudes in Cameroon11, therefore, P. vivax could infect populations in these areas rather than P. falciparum, suggesting P. vivax abilities to infect populations in higher altitudes51. (7) P. vivax may use several receptor-ligand interactions to tightly bind erythrocytes in the absence of a Duffy receptor, e.g., the glycophosphatidylinositol-anchored micronemal antigen or tryptophan-rich antigens52.

Our study had some limitations. Firstly, we identified a limited number of studies reporting P. vivax infection among Duffy-negative individuals. Secondly, we identified high heterogeneity among studies. Thirdly, we observed funnel plot asymmetry which was likely caused by heterogeneity of the ES among studies. Although subgroup analyses were performed, the heterogeneity persisted. Therefore, our results must be interpreted with caution.

Conclusions

Our systematic review and meta-analysis confirmed that P. vivax infected Duffy-negative individuals over a wide prevalence range from 0 to 100% depending on different geographical areas. Future investigations are required to determine if Duffy-negativity is still protective for P. vivax infection.